Jahanara Nissar

While the Street worries about stagnant momentum in the memory markets and expects modest print and guide, we think the highlight of the call today could well be about MU breaking out of NVDA’s grip. Going into the 3FQ24 earnings event in June, we called a near-term top to MU’s fundamentals (link) and cut our $150 price target shortly thereafter (link), before the mid-year sell-off in AI-related semis. Our reason was simple enough – with MU management claiming its HBM capacity was sold-out through 2025 and with little help from the traditional markets, we felt FTM earnings estimates had peaked.

Six month later, and with the stock stuck in neutral, we now see a way out. We think the XPU opportunity is ahead of MU and is yet to price in. As the XPU market heats up, the Street has begun to pay attention and is casting about for AI silicon opportunities outside of the two GPU incumbents. Just as with GPU accelerators, XPU accelerators too need HBM memory.

We suspect US-based hyperscale players may lean towards partnering with MU, instead of the Korea-based memory suppliers, as the HBM supplier for their XPUs. Our checks show that MU may have won the HBM socket at AWS’s Tranium3. Even though we expect this to be a 2026 opportunity, we think the Street has begun to pay attention to AI silicon beyond 2025, as seen in the market’s strong response to AVGO’s Fy27 outlook. Furthermore, we expect incremental HBM manufacturing capacity at MU to come online in Fy26, allowing MU to go beyond NVDA as its sole customer and to diversify into new customers.

We are cognizant of a few additional positives: 1) We think Samsung may no longer be a credible threat to MU’s HBM3e share, 2) the client market, while moribund at the present, is showing modest improvement in visibility into 2025 and 3) geo-political factors, a negative for most semis, nets out of positive for MU.

Going into the previous earnings (4FQ) call we set a tight PT of $110 based on Fy25 EPS (link). Despite significant upside to Nov guidance provided at the previous earnings call, the stock has not been able to break out of the $110 ceiling. With the introduction of XPUs potentially in the mix, we think the Street’s attention now shifts to FY26. As the company diversifies its customer base and expands manufacturing capacity, the potential for new vectors of growth provides avenue for earnings expansion. We are raising our PT to $125, based on 12x to our below-consensus Fy26 earnings estimate. If the XPU opportunities at the various hyperscale players really are of 1mn cluster size each, as AVGO management expects, MU could be looking at significant multi-year opportunities, not constrained by NVDA.

–xx–

The XPU wave is in its infancy: We expect MU management to ride the coat tails of AVGO and blast its way beyond the gravitational pull of NVDA’s GPU-centric ecosystem. That every one of the US-based hyperscale players has been working on internally designed AI accelerators is no secret. MSFT, GOOG and, most recently, AMZN/AWS have rolled out details and high-level road maps. The Street is also generally aware that META and potentially AAPL too are working on internal solutions. However, these efforts have not translated into stock valuations, we think, due to the lack of granularity of the roadmaps and unclear timing of production.

The AVGO earnings presentation last week took a stab at putting a timeline around the XPU opportunity. The surprisingly strong 3-year SAM opportunity that AVGO laid out was translated into significant expansion in market cap. The earnings presentation from MRVL the week earlier too took investors by surprise.

In essence, we think there is great interest in the investment community to learn more about the $value and the timing of AI opportunities outside of NVDA/AMD. Whereas GOOG’s TPU6 has gone into production, delivering revenue to AVGO, the XPUs at the other hyperscale companies are further out in design, let alone production. But the key is this – the Street has begun to put a valuation on the silicon suppliers involved in XPUs. And MU’s HBM is in the sweet spot.

XPUs at hyperscale companies: Just as the ASIC suppliers such as AVGO/MRVL work closely with hyperscale customers during the design phase, we believe memory suppliers get involved early in the XPU design project. This provides HBM suppliers with visibility into the revenue opportunity multiple years away. There are several internal programs ongoing at hyperscale players. MSFT’s 2^nd generation AI accelerators could go into pilot productions late 2025. META and AAPL have internal AI chips in the works, the timing of production though is unclear. AMZN/AWS recently unveiled the 3^rd generation of its internal AI chip.

MU lands AWS’s Trainium3 HBM socket: AMZN/AWS recently announced its roadmap for the 3rd generation XPU chip. We believe Trainium3 is a year away from tape-out. And yet, we believe AWS has already chosen its HBM vendor, due to the need for close collaboration. We believe the HBM contract has been awarded to MU.

We believe that while the previous generation Trainium2/Inferentia2 was a limited release effort to test out the various LLMs AWS has been experimenting with, Trainium3 could go into a wider release. We will not hazard a guess as to the revenue opportunity this socket presents, but it could be meaningful.

For its 3^rd generation, we believe AWS decided to roll out just one ASIC for both training and inferencing purposes. We believe AWS has settled on a limited number of LLMs it plans to take into full production. Trainium3 could become the AI workhorse at AWS in a couple of years. If MU lands this socket, as we think it has, we think MU could be looking as a large driver of growth.

Samsung’s HBM3e may no longer pose a threat to MU/Hynix: Stung by delays, Samsung may have decided to cut its losses and walk away from qualifying HBM3e. We think Samsung may have decided to re-allocate R&D resources elsewhere, such as in the development of HBM4. As such, we think Samsung’s HBM3e is no longer an overhang for MU.

Traditional markets could be stabilizing: The promises at MWC and Computex earlier this year failed to materialize in the AI smartphone and AI PC markets. Memory and flash build-up by suppliers earlier this year in anticipation, instead of selling through during holiday season, ended up in channel inventory. Having said that, the supply chain is setting modest goals for the smartphone/PC markets and adjusting supply accordingly.

Targets for AI smartphones for 2025 have been dialed into the supply chain, setting the stage for improved visibility. Starting off small, less than 20mn units of Android AI smartphone units, it gives a jumping off point for memory suppliers. Minimum requirements appear to be 8GB DRAM and 256GB flash storage. Samsung’s targets for overall smartphones next year too appear to be dialed in. Slightly lower in volume vs 2024 but offset by higher $content.

PCs expectations have been dialed back to pre-pandemic levels. There is very little talk of AI PCs. The focus has returned to traditional metrics such as battery life. Intel is on record saying that its 3nm based Lunar Lake uses MU DRAM. At a recent investor conference, the CFO complained of the high price of MU’s product. We believe Lunar Lake is sole sourced to MU’s lpDDR5x module.

Geo-political factors net out in MU’s favor:

• BIS regulations targeting China data centers – no impact to MU: The Street expects another sets of BIS regulations to be released before end of year, placing restrictions on exports to China of, among other types of semis, HBM modules. We think MU is insulated from further downside due to existing restrictions coming for the Chinese side, regarding the import of MU’s memory products.
• CHIPS Act monies- a positive: MU is a beneficiary of a substantial grant from the US government for use in MU’s planned investment of fabs and packaging facilities in the US
• Political crisis in Korea is driving up cost of MU’s competitors: The Korean Won has dropped ~2.5% against the USD and the Chinese Yuan in the two weeks since the recent political crisis. Korean credit default spreads widened the most among sovereign bonds across the globe last month, second only to Brazil. As we wrote in a recent note, the price of NAND raw dies jumped 2%-3% overnight following the political event earlier this month (link). MU could be an accidental beneficiary of the unexpected price uptick of Korean memory output.

Financials and TP: We are model Fy25 at $38.1bn/$8.2 and with a gross margin on 41% vs. consensus estimate at $38.3bn/$8.96, with gross margin of 43.4%. We believe the Street is modeling gross margin too aggressively in the back half of Fy25. We model FY26 at 45.7bn/$10.47, gross margin 42% vs. consensus estimate of $47.4bn/$13.25, gross margin 48.8. Again, we believe the Street is too aggressive on margin assumptions. Based on a 12x multiple to our Fy26 EPS estimate, we derive a price target of $125. Our previous PT of $110 was based on our FY25 estimate.

Net/Net: We think the XPU opportunity is ahead of MU and is not priced into the stock. We believe MU management has an opportunity to break out of its dependence on NVDA and draw investor attention to new customers for its HBM products. With the Street concerned about near-term dynamics in the memory market, MU management has an opportunity to highlight HBM opportunities over a 2- to 3-year period in the XPU space. We believe there is upside to the stock. We are raising our PT from $110 to $125. KC Rajkumar

The Company filed an 8-k Friday after market close – the NASDAQ has granted its request for an extension for filing the necessary 10-K and 10-Q documents. The company has until 2/25/25 to file the documents and will remain listed on the exchange while it prepares the documents under the supervision of its newly appointed external auditor.

Does this 8-k bring down the curtain on a drama that started three months ago when the company announced a delay in filing its 10-K? Possibly. Three weeks ago, in the face of the Street giving high odds for a stock delisting event, we rolled out a $45 PT and opined that odds of delisting were not quite as high as many thought. Our view was based on belief that SMCI held a uniquely critical position in the all-important business of installing AI infrastructure (link). Delisting the stock would cut off access to capital and result in impaired progress in AI data center build out. We also opined that the odds of alternatives such as DELL picking up share from SMCI were not quite as high as many thought. We called for the run-up in DELL on investor expectation of significant gains in AI infrastructure to fade (link).

In the past few week SMCI picked up steam as positive headlines began to hit the tape. With the 8-k filing on Friday, the stock cut above our PT in after-hours action. We are now raising our PT from $45 to $60 based on the expectation that, with the NASDAQ decision in hand, many investors are likely to return. Some brokerages which dropped coverage are likely to reverse their decision.

Are there knotty legal issues involved with the delay in filing? We will not know until the agencies rule on the matter. There is no guarantee that the company will be able to file the documents by the appointed date to the satisfaction of the external auditor and the stock exchange. However, for the next 2+ months, until the filing deadline in February, the event-driven volatility is likely to wind down while investors turn their focus to company fundamentals. And that is reason enough to raise our PT.

–xx–

In the recent 3-month period of extreme volatility, investors had an opportunity to pick up the stock at deep discount. Many on the Street were doubtful whether the company could avoid de-listing. In anticipation of delisting, investors believed that AI infrastructure build-out could continue with the involvement of other players such as Dell and Foxconn. We note that SMCI has had nearly 100% share of the liquid-cooled GPU server market in the US.

Event-driven call: Three weeks ago, we made a call that SMCI’s role in the industry was central enough that the AI ecosystem would be forced to come together and find a solution to SMCI’s regulatory woes. Delisting would trigger potential loss of access to financing and incur inability to participate in the capital-intensive infrastructure business. We made a call that the regulatory authorities would give SMCI the benefit of the doubt and allow it to remain listed while the company prepared the 10-K/10-Q documents (link). With this in mind, and with the stock trading at ~$21, we rolled out a PT of $45 based on 15x to our below-consensus FY25 estimate of $21.7bn/$2.93. Soon thereafter, the company announced the appointment of an external auditor. We reiterated our $45PT call (link).

Fundamental call: Besides the event-driven call, we made a fundamental call as well. We called for the run-up in DELL into its earnings event to fade. Investors assumed that the revenue shortfall reported by SMCI at its Q1 earnings even would show up as gains at DELL. It didn’t. DELL too missed its Q4 revenue expectations by over a billion dollars, in part due to its inability to ramp Blackwell shipment. Dell management said Blackwell orders had gone into backlog. However, DELL’s backlog increased by less than a billion $s q/q, hardly enough to signal it had picked up the baton from SMCI. DELL sold off hard after earnings. HPE too, at its recent earnings call, did not indicate it was stepping in to fill a potential hole being left behind by SMCI.

Raising price target: We are raising our PT from $45 to $60 based on 20x to our unchanged FY25 eps estimate. With the prospect of delisting now fading further, we expect many funds to return to the stock and some of the brokerages to resume coverage. And secondly, the recent earnings results at DELL and HPE show that they are not likely to pick up the slack if SMCI were to fall by the wayside. We think it may have become apparent to investors that SMCI holds a unique position in the Gen AI infrastructure industry, in the business of setting up data centers with tens of thousands of liquid-cooled GPUs at scale.
KC Rajkumar

In a disappointing presentation at an investor conference yesterday, the CFO, newly appointed as interim co-CEO, missed an opportunity to wipe the slate clean and start afresh. If, at the very least, Q4 guidance was reiterated, that would have been progress. But the vague language leaves investors guessing. Progress on 18A? Looks like the schedule has slipped, thus confirming worries on the Street. External customers at 18A? No mention of. Concrete plans for use of the just-delivered monies from CHIPS Act? None provided. Vague promises of improving IFS margins next year, for which we suspect there are few takers on the Street. The icing on the cake for short sellers was the lowering of Product margin outlook.

Even though there is every reason for the stock to trade down from current levels, we suspect short sellers would be wary of headline risk. If the Board were to announce an acceptable permanent CEO, no one wants to be on the wrong side of a SBUX-type move. In case of INTC though, none of the candidates mentioned in the media seem all that exciting. And it is doubtful whether elevation of either one of the two interim co-CEOs would do the trick.

Even if a credible CEO were to be named, given the daunting challenges faced by the company and the sore disappointment the initial excitement around the appointment of the previous CEO turned into, we think an initial pop in the stock would only encourage short sellers.

Despite the dysfunction in the C-suite, the company still accounts for ~70+% of PC and server unit sales. Despite the challenge from AMD and ARM, we think it safe to say Intel is likely to retain majority share of global PCs and server CPU into the foreseeable future. There is value in that. But few investors can step in as the slide in financial metrics continues. And so, the stock remains stuck in purgatory, somewhere between book value and tangible book value.

But hope springs eternal. The new CEO appointment may turn out to be a clearing event after all, and with it, new investment from private sources to follow. To the growing list of potential CEO candidates out there, allow us to throw another one – Jim Keller of Tenstorrent. Intel is one of the last remaining repositories of silicon engineering talent in Silicon Valley. There is value in retaining that talent under one roof. We suspect that talent pool is deeply demoralized. It would take a technology leader with impeccable credentials to re-invigorate the workforce. More cost cutting just isn’t going to do it.

–xx–

Q4 guidance update– wishy-washy: The CFO’s comment ‘we stand by the guidance we gave at earnings’ simply does not cut it, in our view. If the idea was to reiterate guidance, then it would have been appropriate to include a statement in the 8-k filed two days ago announcing the departure of the previous CEO. Or at the very least we would have expected the CFO to say something along the lines of ‘the quarter is tracking to our expectations’. In the absence of such a statement investors could assume the quarter is likely tracking BELOW expectations.

Intel 18A update – further delay: Four months ago, at the Q2 earnings call, 18A schedule was set to volume ramp in 1H25. A month ago, at the Q3 earnings call, there was a slide in the language; 18A was simply characterized as ‘progressing well’. Yesterday at the investor conference, the goal post was moved yet again – volume ramp is now expected only in 2H25. This means that Intel’s 18A products, Diamond Rapids for servers and Panther Lake for PCs have slipped by an additional six months at the very least.

IFS external customers – slim pickings: The Ops Chief’s comments seem to confirm what investors had come to suspect – there are no major external customers at 18A. And no real commitments for the next generation 14A either. Pat’s dream of Intel offering foundry services to fabless companies now seems to slip further away. And if IFS has no credible external customers, what would be the justification in spinning IFS out as a separate company?

Product division – slipping margin: The CFO appeared to talk down 2025 margin expectation of the 3nm-based client product Lunar Lake. While reason given was high memory costs, we suspect the real reason is the wafer price increase out of TSM. Next generation client product Panther Lake, based on 18A is expected to improve product margin as more chiplets are insourced. However, with 18A production ramp now apparently slipping by from 1H25 to 2H25, we suppose the ramp of Panther Lake in IFS slips from 2H25 to 1H26, thereby delaying margin improvement.

While the above is a 2025/26 narrative, more worrisome is the CFO’s comment on margin slippage over a multi-year basis, from the low 50s to ‘something with a four handle on it’. With IFS margin deeply in the negative territory, Pat’s hopes of margin recovery back into the 60s territory now seems further away.

Capex – feels like a cut: In his role as co-CEO, the CFO says he has committed to the BOD for extracting incremental ROIC from investments already made. It feels like he is making a cut to the 2025 capex outlook he had provided at the Q2 and Q3 earnings calls – gross and net capex of $20bn-$23bn and $12bn-$14bn respectively. With 18A likely delayed by six months, there is justification for pushing out capex ramp for 18A capacity.

Taiwan supply chain partners – left wondering: In the final month of his tenure Pat had made well-publicized trips to Asia, we suppose to re-establish relationships with supply chain partners in light of the soon-to-be consummated CHIPS Act grant approval. His abrupt exit we think may have rendered those relationships in disarray. Price commitments and payment terms he may have worked out with key Asia partners may be scrapped. Make no mistake, there is a cost to Pat’s abrupt exit. We doubt if either of the two co-CEOs would be able to fill his shoes.

Net/Net: Even though there is every reason for the stock to trade down from current levels, we suspect short sellers would be wary of headline risk. If the Board were to announce an acceptable permanent CEO, no one wants to be on the wrong side of a SBUX-type move.

Performance and financial metrics appear to have deteriorated further. Pat’s abrupt exit weighs on company morale and supply chain partner relationships. We expect the stock to remain in purgatory awaiting announcement of a new CEO.
KC Rajkumar

South Korea casts an outsized shadow on the global electronics supply chain. The shocking series of events in S Korea over the past 24 hours, a country known for its political stability, is sending shivers down the supply chain in Asia, already on edge due to the trade uncertainty radiating from the US. MU may be an accidental beneficiary.

–xx–

The Won lost: The Korean Won was down as much as 2.8% overnight after the surprise declaring of martial law by the President. Over the next 24 hours as the lawmakers of both parties voted down the President’s declaration and after the President lifted martial law, the Won recovered to down 1% from its pre-martial law level. We note that forex shocks on the order of 100bps is hugely problematic for global supply chains.

Cost of imports: Even if Korean politics return to status quo, the damage to the credibility of the nation’s currency may linger. Overnight, the Japanese Yen strengthened against the Korean Won. The Japanese Yen may matter more directly to the Korean economy than the USD. Cost of imports from Japan, a major supplier of raw materials to Korea’s silicon fabs, is likely to spike up. And as a result, the price of silicon chip exports out of Korea too is likely to move up.

Feedback from the Taiwan supply chain suggests that there is marked concern. The prices of commodity memory products in Taiwan reacted almost immediately. While it may take days for prices to settle down, we are hearing commodity NAND flash prices jumped 2%-3% overnight. The abrupt jump in flash price is an added burden to PC ODMs, whose margins run typically below 10% and are faced with a weak demand environment.

MU – an accidental beneficiary: Upward movement in Korean memory prices could be a positive for MU, especially as its USD denominated input prices are relatively insulated from the Won volatility. On the flip side, currency fluctuation could have knock-on effect of suppressing end demand for consumer goods, which may reduce the demand for MU’s memory products.

Trading call: A stabilization in falling memory prices could be positive for MU. And WDC. We maintain our $110 PT for MU.
KC Rajkumar

NVDA stock traded up ~20% since early October when the CEO started a media/Street campaign to message that Blackwell demand is ‘insane’ and that ‘everything’s on track’. With the stock close to its all-time high, we detect investor nervousness given that the valuation is leveraged to Blackwell, a future product cycle that has had trouble getting off the starting block due to well-known heating issues. We will wait for an update from management regarding the progress with the fix it claimed to have accomplished.  

Having said that, the fact remains that NVDA is the only game in town. Going into the recent AMD AI event, we said we see little out there which makes us believe AMD could make a dent (link). At hyperscale players, we do not expect their internal silicon to make a dent for the next couple of years. An exception is Google’s TPU program. Even though Google Gemini could be a price spoiler, Google’s AI strategy is yet to emerge. Nevertheless, inference token cost seems to be coming down on its accord, without much pressure from Google, and with it, the pricing of Hopper GPUs.

NVDA management is tasked with having to convince investors that it has visibility for Blackwell demand at least into 1FH26. Colorful phrases may not be enough. Phrases such as ‘demand is outstripping supply’ may not mean much if supply is constrained due to yield issues. Are customers willing to plunk down ~$1bn-$2bn in upfront hardware cost for a 20K GPU datacenter without first gaining conviction that the problems associated with the initial delay in launch have been resolved? We do not think so.

Then there is the matter of demand for Hopper. At the previous earnings call management set the expectation that Hopper growth was to continue, even as Blackwell ramped in 4FQ25. Going into this call, there is some worry on the Street that Hopper growth is in question. Falling price of Hopper server rental and token cost are now being noted on the Street. We focused on this issue three months ago in our July quarter preview (link). Besides, could it be possible that the giant miss of print/guide at SMCI was in part related to falling demand for Hopper? We think so.

Going into the previous earnings event, with the stock trading at an all-time high, we suggested that investors may step aside and look for a better point of entry. After the earnings call, the stock dropped ~20% before finding its footing. We are cautious going into this earnings event as well – step aside and look for a better entry point.

As the Gen AI cycle matures, the initial excitement wears off, and hyperscale customers put more focus on ROI metrics, investors need to anticipate growth rates inflecting down. Even a year from now, consensus estimate for data center revenue growth in Oct 2025 is 50% y/y. Is that realistic? We do not think so.

–xx–

Blackwell’s lead customer – NVDA itself? According to AWS blog and media reports, the Ceiba project at AWS is a joint development effort with NVDA. The project is based on a purpose-built liquid-cooled GPU datacenter large enough to house 20K GPUs and is expected to be among the first large customers of Blackwell GB200. NVDA’s Blackwell revenue in the Jan quarter we suspect will largely be derived from AWS/Ceiba. We believe Foxconn is the ODM chosen for this project. We expect the 20K shipment to complete in 4FQ/1FQ.

The anchor customer of the data center may be NVDA itself! We believe the shell construction may be externally financed. The hardware infrastructure too is probably financed by third parties. As such, we believe AWS’s capex exposure and risk of underutilization could be minimal. We believe subsequent customers of Blackwell could hold off orders for Blackwell until NVDA delivers performance metrics from the AWS/Ceiba data center.

When does Blackwell order from MSFT kick in? Microsoft runs some of the largest LLMs in the nascent AI industry. We expect Microsoft to be a key customer of NVDA’s Blackwell, given the inferencing needs of GPT-4 class of models and the training/fine-tuning needs of Orion class of models. As such it is surprising to us that MSFT does not appear to be the lead customer of Blackwell. We wonder if MSFT is waiting for performance data from AWS/Ceiba before committing to Blackwell for its data centers.

Sovereign AI? Beyond AWS/Ceiba delivery in 4FQ/1FQ, and perhaps Microsoft in 1FH26, it is not clear to us where would Blackwell go. NVDA management often points to potential demand from sovereign customers – India, the Middle East etc. The timing of delivery is not clear to us. And are these even credible customers? Their business cases are less than obvious, the timing of orders tied to political dynamics, especially as the US government gets more stringent with export rules.

SMCI earnings disappointment – a view into Hopper demand? At its recent earnings call SMCI provided preliminary Sept quarter revenue at ~$6bn vs. the original guidance of $6bn-$7bn, a $0.5bn miss at midpoint. The company guided December quarter revenue to a $5.5bn-$6.1bn vs. consensus of $6.9bn, a $1.0+bn miss at midpoint.

Management explained their giant miss as due to non-availability of Blackwell. However, this explanation is inconsistent with management expectations set three months ago at the June quarter earnings call. At the June quarter event Management explicitly stated that it expected no revenue from Blackwell in the September quarter and very little in the December quarter. It seems to us that management’s explanation set forth at the September quarter earnings call lacks credibility.

Investors seem to take the view that the miss to both quarters was due to end customers re-directing orders away from a company that appeared to be on its way to getting delisted. We can understand customers trimming order pipeline in out-quarters, say in March/June quarters. But why would customers cancel firm orders for delivery in September and December, as the company appears to be well capitalized to handle working capital requirements. If the demand for Hopper is as strong as NVDA management says it is, why would data center customers cancel orders?

Our view: While there may be many factors that may have gone into the mix, it seems to us that softening in demand for Hopper could be a key reason for the SMCI miss, either due to lack of end demand from customers of data centers, or due to non-availability of liquid-cooled data center shell space. Reduced demand for Hopper at SMCI naturally implies reduced demand for Hopper and networking gear at NVDA.

With Blackwell expected by NVDA management to amount to ‘several billions’ in its Jan quarter, the overwhelming majority of the ~$37bn worth of Jan quarter revenue consensus expectation comes from Hopper. Any softening of Hopper revenue could have meaningful impact to NVDA’s Jan quarter guidance.

Demand centers in Asia: In the 6-month period ending in July 2024, NVDA reported the revenue exposure to Taiwan, Singapore and China geographies at 46%, just as large as the exposure to the US. However, Taiwan and Singapore do not have giant data centers. Are GPUs shipped to these two countries finding a home in Europe? If not where? With the US government clamping down on shipment to China via third nations, sales into Asia geography, especially Singapore (11% revenue exposure) could come under regulatory scrutiny.

Taiwan ODMs may absorb surplus H100: Media reports of H100 being made available to Tier 2 ODMs such as Asus, ASRock and Gigabyte implies demand shortfall in the US vs expanded supply, in our view. We are not clear who could be the end customers of these ODMs. We do not think they are supplying customers in the US or Europe. We are not aware of any meaningful data centers in Asia outside of China.

Outside of Foxconn, none of the Taiwan ODMs have expressed an expertise in liquid-cooling technology. Given the advantages posed by liquid cooled racks, going forward why would data centers invest in air-cooled GPU racks? It appears to us that Tier 2 ODMs may simply act as holders of channel inventory, rather than active suppliers into working data centers. As such, sale into Tier 2 ODMs could obfuscate true end demand.

Net/Net: We are cautious into print due to our view 1) demand for air-cooled H100 could be softening, 2) the uncertainty surrounding the performance of Blackwell could put the brakes on demand after the initial rush of orders and 3) limited availability of liquid-cooled data centers could crimp GPU demand.  

On a longer-term basis, as the Gen AI cycle matures, the initial excitement wears off, and hyperscale customers put more focus on ROI metrics, investors need to anticipate growth rates inflecting down. Even a year out into Oct 2025, consensus estimate for data center revenue growth stands at 50% y/y growth. Is that realistic? We do not think so, if not for any other reason but for the well-known issues with power/water considerations. We think growth expectations must reset for the stock to find a firmer footing.

KC Rajkumar

Versus the muted expectations on the Street for near-term iPhone momentum, we are long into print. The Street is inundated with chatter of order cuts at iPhone16. We take a more nuanced view and look for upside to Street’s lowered expectations for Dec quarter.

Our expectation for Dec upside is not based on a boost from Apple Intelligence. Rather, we focus on more traditional dynamics. Our view is based on the potential opportunity Apple may be seeing in the general weakness at Samsung, unrelated to the relative progress either of the smartphone players have made in AI.

Signs of Samsung weakness are out there, and yet the Street has not picked up on them; not just in the weak qualitative outlook Samsung provided for its smartphones at its earnings call earlier today, but also in QRVO’s disastrous guidance two days ago. Our work into the supply chain shows Samsung potentially giving up shelf space this holiday season at telcos and retailers, an opening we believe Apple would be only too willing to pounce on. Instead of Apple, we think it is more likely that Samsung has cut build plans for the holiday season. And that leaves Apple with a competitive opening.

Following Apple’s WWDC event a few months ago we took the view that Apple Intelligence, which we are bullish over a multi-year basis, would be no more than a moderate catalyst for iPhone sales in FY25. Into the June quarter earnings call, Apple bulls were looking for iPhone growth in the teens for Fy25 and overall revenue up double digits vs. our estimates of both up mid-single digit (link). Apple’s tepid guidance for the September quarter and the negative news flow over three months whittled Street expectations down to more realistic levels, now inline with our estimates a quarter ago.

Over the past three months, AAPL has traded only inline with the market, justifying our call into the previous quarter earnings event to sit out the June quarter earnings event. Into the earnings call today though, given muted Street expectations, we would be long into print. FQ1 iPhone guidance inline with average consensus of up mid-single digit may be enough to lift the stock.

We remain comfortable with our long-term view that the hidden value in the stock lies in its potential to pull up to par with AI peers over time but without the extraordinary spending on opex/capex at peers. At a slightly reduced multiple of 32x to a slightly higher FY25 eps estimate vs a quarter ago, we maintain our $240 PT (link). However, depending on the tenor of the call today, we could be looking to raise our PT.

–xx–

The Android ecosystem runs into headwinds: Our work into the smartphone channel shows Samsung potentially cutting build of its high-volume smartphone models, a rather odd finding given that Samsung’s A-series models have become a mainstay on shelves in the high-volume pre-paid cellular market. When QRVO, at its earnings call two days ago, talked of its largest Android customer cutting back its mid-tier models (the A-series at Samsung) and tilting towards entry level model, we think we have confirmation of our suspicion.

This trend towards ‘bar-belling’ smartphone models, i.e. terminating mid-tier models and allocating volume towards the entry-level and premium ends of the spectrum has been going on for 1-2 years. However, in recent months, for reasons we don’t quite understand, this trend appears to have accelerated at Samsung. The latest IDC report for Q3 shows Samsung shipment down mid-single digit vs. overall global  smartphone units up mid-single digit. Apple our performed Samsung – iPhone unit growth was reported inline with the overall market. Apple picked up 200bps to share q/q, at the expense of Samsung and Xiaomi, according to the report.

China shipments for Apple too seems benign, with the Canalys report showing iPhones holding up at ~10mn units every quarter for the first three quarter of every year for multiple years, and a seasonal uptick to the high teens in CQ4.

Signs of stress at Samsung, Apple likely to exploit: Across the gamut of its product segment, Samsung appears to demonstrate weakness and a loss of traditional leadership – logic fabs, HBM innovation, consumer goods and mobile phones. We find that a general weakness at Samsung is having observable effects on its upstream/downstream partners. Into the AMD print a few days ago, we anchored our negative thesis on, among other reasons, Samsung’s HBM woes (link).

In the mobile space, we believe Samsung is pulling back on its shipment plans into the holiday season. At its recent earnings call, QRVO said revenue at its largest Android customer was less than that anticipated earlier this year.

We think Apple is likely to exploit the pullback in Samsung’s positioning in the retail space by potentially building additional iPhone volume for the holiday season. We expect Apple management to signal more optimism for iPhone in the Dec quarter than it did for the Sept quarter.

Meanwhile on the AI front, Apple ploughs on as the Android ecosystem hesitates: We think Apple’s M4 chip does not have the horsepower required to perform the kind of Gen AI functions that is likely to make Gen AI a selling point for Apple devices. On the other hand, AI SoC at QCOM and MediaTek appear to have more raw horsepower than Apple M4’s 38 TFLOPS. Both in terms of TFLOPS and the sophistication of the o/s, we think the Android universe has an edge over iPhones.

However this edge Android currently has over the Apple universe could soon disappear as 1) the Android universe hesitates pushing its advantage; Samsung, Google and QCOM seem unable to come together and deliver eye-popping solutions and 2) the vertically integrated Apple ecosystem, with a line of sight to 2nm silicon next year has a shot at better integrating hardware/software and vault over the present-day advantages the Android ecosystem may have.

We think major improvements in image/video generation in iPhones can be expected only in the FY26 cycle when 2nm M5 silicon launches.

Apple – patient and prudent: We expect investors to complain about the lack of eye-popping AI applications on the current iPhone16. However, Apple’s hyperscale peers, despite their heavy investment in opex/capex find themselves unable to impress investors, witness the dour response to META and MSFT earnings last night. Do investors really have a choice but to give Apple the benefit of the doubt and hope that Apple extracts superior ROI from its prudent, one might even say skeletal AI investments?

Financials: We are leaving our model relatively unchanged from a quarter ago. We continue to model iPhone segment and the overall revenue growth up 6% and 6.7% respectively. Street estimates have drifted down to where we have been, as the more bullish expectations got whittled down. We have been modeling Fy26 revenue at $443bn. Three months ago, the bulls got to our F26 estimate in Fy25 itself! Not anymore. Expectations for FY25 have moderated even as the stock has held up inline with the market.

At a slightly reduced multiple of 32x to our slightly higher FY25 eps estimate, we maintain our $240 PT. However, depending on the tenor of the call today, we could be looking to raise our PT.

Exhibit 1: Key financial estimates

Source: Bloomberg, Lynx Equity Strategies

KC Rajkumar

An article in Digitimes highlights large-scale problems in Samsung logic foundry operations, not very surprising given the negative outlook from ASML last month. This piles on top of the series of negative comments outlined in Samsung’s earnings call last week – DRAM/NAND capacity adjustments to normalize inventory, ongoing slowdown in smartphones, lackluster progress in HBM3e.

We have been highlighting the widening impact of Samsung’s woes to companies in our coverage. A severe drop in overall profitability, as reported at the 3Q earnings call, we believe is having an impact on Samsung’s ability to carry out core functions, regardless of the tenor of end demand. In this note, we briefly highlight the Samsung-related calls we made last week to companies in our coverage.

Positive for Micron: We wrote in a note last week, our checks showing at least one DRAM wafer fab at Samsung has been taken offline since early last month (link). Into MU’s earnings in September, we wrote Samsung had begun to cut back packaging of DRAM dies (link). Furthermore, continued delays in Samsung’s HBM3e qualification holds in check HBM supply into the channel. Samsung’s actions curtailing the supply of DRAM and ongoing delays in HBM3e are clearly beneficial to MU. While we have no expectation for MU stock to head for its highs for the year, we think Samsung’s woes are a tailwind to MU. We think there is upside to our $110 PT., but likely no more than $120.

Negative for AMD: We made a negative call into AMD print last week. We wrote in our preview, Samsung’s inability to deliver HBM3e as a key reason why AMD may be unable to provide upside surprise to its MI300X outlook (link). Samsung is a key supplier of HBM to AMD. Management raised 2024 outlook only inline with consensus and disappointed the Street by not providing a peek into 2025. The stock is down ~15% since the earnings event and near the lows of the year, as we had called out in the preview. We do not expect buyers to step in, as there are few positive catalysts to lift the stock.

Positive for AAPL: Worsening profitability at Samsung may be impacting its balance sheet, which in turn impacts its ability to finance smartphones for telco and retail channels. We think this works to AAPL’s advantage. We think iPhones, regardless of the progress in the rollout of Apple Intelligence, are likely to grab incremental shelf space as Samsung recedes from its dominant position in the US and Europe markets (link). In the haze of calls on the Street for cuts to iPhone build, we think the Street underappreciates the potential for weakness at Samsung’s financing muscle to provide tailwinds to iPhone supply into the channel. There is no weakening of Apple’s financing muscle. We expect AAPL stock to gain momentum as the Street digests the competitive landscape. We maintain our $240 PT.

Net/net: We do not think the Street appreciates the widening impact of Samsung’s woes on the competitive and supply chain landscape. This is especially so in the case of AAPL iPhones.

KC Rajkumar

A darling of Wall Street until just a few months ago, SMCI now finds itself in financial, regulatory and legal purgatory. When we took a negative view on the stock into the previous earnings call, we could not have anticipated accounting problems and the existential threat the Company now faces (link). Into the previous earnings reports we expected the Company to guide revenue soft. Instead, the Company raised revenue outlook significantly, but also missed profitability significantly. Investors were not amused.

With the Company now staring into the abyss, what does it do from an operational perspective? Company management may avoid brightening up short-term optics and instead focus on mending that which ails it. And what ails it? In getting into the liquid cooled GPU server market requiring high investment, we think the company has bitten off more than it can chew.

Inventory on hand exploded to $4.4bn exiting Fy24, tripling from a year ago vs revenue over the year merely doubling. The abrupt hit to inventory arising from high cost of liquid cooled kits for H100/H200 directly impacted margins and profitability. Cash flow from operations for Fy24 was reported negative $2.5bn vs. positive $663mn a year ago. Over the course of Fy24 the company raised ~$1.6bn in a convertible issue; a secondary offering raised $1.75bn in gross proceeds. And yet the Company exited the Fy merely neutral on a net cash basis.

In order to ensure long-term survivability, we think it has no choice but to remove the even more expensive Blackwell program off its shipment pipeline. And this necessarily means lowering Fy25 revenue guidance. Recall that the Fy25 revenue guidance of $26bn-$30bn includes revenue from Blackwell in 2FH. We think the Company’s weakened financial situation leaves it with few alternatives. Could this improve the trajectory of gross margin in the back half of the FY? It could. And that might turn out to be positive for the stock.

While we have no view into the specifics of the accounting problems triggering the Company’s external auditors to resign, we wonder if management’s response to abrupt changes to the cash flow and balance sheet items may have had something to do with it.

–xx–

The high toll of GPU servers: Even before the abrupt drop in margin reported last quarter (Exhibit 1), Company fundamentals had already begun to deteriorate. Inventory on hand began spiraling upwards (Exhibit 2) and y/y growth in opex inflected sharpy upwards (Exhibit 3). Recall that shipments/inventory of NVDA servers have been associated with the Hopper family of GPUs. Transition to the Blackwell family is likely to raise inventory and cost burden further, not just because of the higher upfront GPU cost but also due to the higher component costs associated with higher power and more complex cooling systems.

Unless SMCI does another round of capital raise, the balance sheet may not allow the additional burden of taking on Blackwell. Given the current regulatory situation, capital raise is quite out of the question.

Our view: We think SMCI may have to give up its ambitions for Blackwell servers. We expect Foxconn to take over as the lead (and perhaps the only) ODM for Blackwell. Perhaps even a portion of SMCI’s Hopper pipeline may have to be given up, to right-size the balance sheet and cash flow to the new reality.

Medium term – focus on H100: If the Blackwell program comes off the books, we think SMCI’s prospects have a better chance of mending. We may be in the minority in this view, but we think H100 servers could well become the workhorse for AI inferencing well into the future. Whereas Blackwell and its follow-ons could be the workhorse for modeling workloads, we think H100 could remain the choice for inference workloads due to rising availability, falling prices and a growing secondary market. As such SMCI may not lose much by giving up on its Blackwell ambitions and instead focus on the H100, a chip that SMCI has had quite some experience with. 

With time and experience, SMCI could become more efficient in sourcing, building and shipping liquid-cooled H100. We think SMCI has a better chance of mending its inventory, opex growth and profitability if it sticks to the H100. We think demand for H100 servers is not going to sunset with the arrival of Blackwell. The two GPUs run complementary workloads. We do not expect Blackwell servers to displace H100 servers in AI data centers.

Expertise in liquid cooling systems: We find that the state of the art of liquid cooling systems for GPU server racks is surprisingly primitive. Our view is that NVDA has not been closely involved with the design and manufacturing of the liquid cooling systems. The ODM/OEM vendors have gone ahead and largely designed their own system, resulting in a hodge-podge of designs across server rack vendors. We have heard from the downstream channel that the poor design of cooling systems leads to temperature variance on the GPU board, non-optimal operations of servers and workload inefficiencies.

Among the vendors, our checks show SMCI delivers some of the better designed systems. In working closely with NVDA, we think SMCI has built up internal expertise in designing liquid cooled systems. Innovations in this arena could offer SMCI an avenue to increase customer satisfaction and profitability.

Net/Net: SMCI finds itself facing daunting challenges. The first task at hand is to mend its unsustainable trajectory of inventory and profitability. And this requires, in our view, retrenchment of its ambitions. Given its attenuated ability to raise more capital and dire financials, we think SMCI may have to lighten up on Blackwell ambitions in order to better serve the H100 market and to ensure long-term viability of the Company.

KC Rajkumar

Semis had been sailing smoothly for a little over a month until yesterday when the sector ran into a spot of rough weather. Back-to-back squalls had the Street running for cover. Two separate headlines drove semis – ASML down 16% and NVDA down 4% dragged the sector down and forced the bulls to revisit their thesis. Rather than go for a diffuse over-arching negative thesis, we’ll attempt in the following our thoughts on pointed reasons for the headlines and how to trade the volatility.

From a trading perspective, 1) we look to buy the sell-off in AMAT/LRCX following the ASML earnings release, 2) we would not dismiss the NVDA-related media story yesterday speaking to country-specific export caps, we think this story could develop further resulting in NVDA coming under pressure and 3) as for TSM, until the AI export policies are clarified we think it prudent to stay on the sidelines.

–xx–

ASML – what happened? In an unexpected pre-release of the earnings packet, the company lowered its revenue outlook for 2025 from €30b-€40bn to €30b-€35bn due to 1) largely expected lowering of revenue contribution from China to 20% vs. 2024 contribution running significantly higher and 2) of much higher concern to the Street, demand push-out of EUV tools due to slower ramp of new nodes at ‘some’ logic foundry customers. No prizes for guessing who is being referred to as ‘some’.

ASML – China export risk – our view: There is some concern in the semicap circles that in the waning days of the Biden administration, more onerous export restrictions could be imposed on the industry. Having said that, we think ASML has had incremental China risk vs. US-based peers. Based on statements from AMAT and LRCX, we think US-based firms have already de-risked export control risks in large measures. Non-US based vendors such as ASML and TEL have pushed back against US pressure for much of the year. Until now. A month ago, it was reported in the media that the Dutch government had imposed new export controls on ASML thereby aligning itself closer to US export policy. We think the large-sized cut in ASML’s 2025 China exposure disclosed by the company yesterday is in response to the new export policy handed down by the Dutch government.

Meanwhile, AMAT and LRCX have already been hewing closely to US government policy, and therefore their incremental risks to further export restrictions may be limited. In this regard, the sell-off in AMAT and LRCX yesterday may have been excessive.

ASML – Advanced logic risk – our view: We think the ‘delayed timing of EUV demand’ is related to INTC. Should this be a surprise to investors? We have been down this road after Intel announced a capex cut at its Q2 earnings call.  We take the view that a significant portion of the ASML miss is due to Intel delaying EUV demand. And just as we did the last time around (link) we will take the view that outside of ASML, the other major semicap names had already dialed out contribution from INTC. We think AMAT/LRCX offset weakness at INTC by strength at TSM wafer foundry and packaging.

Semicap – Net/Net: After the vol subsides, we will be buyers of AMAT and LRCX. But just so that there are no further surprises, we would wait for a few more weeks and get past the US election, before backing the truck up.   

NVDA – what happened?   Bloomberg put out a negative article pre-open, causing NVDA and AMD to sell-off at market open. The article, quoting unnamed sources, reported that the Biden administration was looking to cap sales of advanced AI chips from NVDA and other companies on a ‘country-specific’ basis, to countries beyond China.

This is not the first time this such an idea has made it to the tape. On previous occasions, due to the lack of follow-through from official government sources, the Street shook off the news. However, this time we would not be quite so sanguine. We would not rush in to buy the dip. If confirmation were to emerge from official government sources NVDA could see further downside.

That there is a common theme to the export controls the US government is seeking to slap on export of semicap equipment/consumables and of advanced AI chips cannot be denied – both categories of export controls have a national security angle to them. In the waning days of the Biden administration, it is possible that government agencies are making one last ditch effort to leave a permanent impact; it would be months before the next administration would be in place to make a similar effort, if at all.

NVDA – some questions: We think the key to framing the downside risk lies in assessing which countries could fall under the export capped status. If the targets countries are the Middle East countries mentioned in the Bloomberg article, there may not be much to worry about given NVDA’s low exposure to said countries. But what if countries ranking higher on the revenue exposure scale are under review by the government?

As per NVDA’s latest 10Q, on a 6-month basis, Taiwan and Singapore each account for ~17% vs. the US accounting for 46%. Nearly 90% of NVDA sales in the two most recent quarters are in the data center sector. One infers there is a whole lot of data center GPUs headed to Taiwan and Singapore, countries which do not have the kind of export controls China is subject to.

Combining the revenue contribution from Taiwan, Singapore and China adds up to the contribution from US-based customers. Unlike commodity CPUs, NVDA’s AI GPUs are mostly done direct-to-customer. There isn’t much of a mystery as to who are the direct customers in the US. But who are the direct customers in Taiwan and Singapore? Are there new data centers being built in either of these two countries?

NVDA – Net/net: Is it possible that the countries which could face export caps for advanced AI chips are Taiwan and/or Singapore as opposed to countries in the Middle East the Bloomberg story seems to speculate? If the US government could place pressure on friendly nations such as The Netherlands and Japan to prevent technology leakage of silicon fab equipment, why couldn’t the US government place pressure on friendly nations such as Taiwan and Singapore to prevent leakage of advanced AI chips to China? If indeed it is Taiwan and/or Singapore that the US government is considering putting export cap on, then the risk to NVDA’s revenue could be considerable. Until the identity of the countries facing export caps is understood from official sources, we think it may be unwise to step in and buy the dip.

TSM – thoughts into print: The narrative out of TSM is largely anticipated 1) further pull-in of CoWoS capacity spending in order to support extraordinary demand from AI and 2) unchanged and muted demand from end markets outside of AI. The stock ran up into print in the expectation of 2024 revenue beating the previous guidance ‘modestly above mid-20s’ now running perhaps closer to 30%.

In order for the stock to make a new high, TSM management needs to 1) provide a qualitative view into 2025 and/or 2) raise the 5-year for AI revenue above the 50% CAGR outlook provided at the March’24 earnings call.

TSM – Net/Net: Rather than TSM being a bellwether for semis like it used to, given the outsized influence of NVDA, TSM may be under the gravitational pull of NVDA. Negative headlines for NVDA, like the one yesterday, are likely to weigh on TSM as well. While we are comfortable with our NT$1250 PT, we are not looking to add to position until there is some improvement in the traditional end markets. For now, we will stay on the sidelines.

• KC Rajkumar

Google hosted an event to roll out Gemini Nano powered applications on the just launched Pixel 9 smartphone. The full set of AI features to be made available in Android models to launch later this year.

Quick summaryGoogle rolled out a series of small but significant improvements in existing Android features powered by Gemini Nano, some new features, all of them available on Pixel 9.Nothing jumps out as a killer app, but that perhaps was not the intent.Google’s intent we suspect has been to make the overall experience more appealing – and make you switch from Apple iPhone.Many Gemini features available on existing Android models.The full experience of Gemini Nano is on Android 15 to roll out later this year.The AI features demonstrated today – many of them can be tried out today on the just launched Pixel9, says google management.

Apple has some catching up to doWe were aware that Apple had not rolled out SDK for Apple Intelligence at WWDC.As we wrote in our June 16th note where we raised Apple PT to $240, we noted iPhone16, at launch, would have AI apps mostly built internally at Apple; 3rd party apps to come later.According to media reports and confirmed later by Apple management, AI capabilities to be rolled out to developers ‘in waves’ during Fy25, and not prior to launch of iPhone16.Which is why we did not model our Fy25 iPhone est too aggressively. We think AI upgrade cycle at Apple is a multi-year cycle, akin to the rolling improvement in iPhone camera over many years.Apple’s valuation is based on more than just the number of devices it sells. Its valuation is justified by its decision to not use NVDA chipset, thereby sidestepping expensive infrastructure costs and the delay issues that now bedevil the likes of MSFT.

Google vs. AppleEven though, at launch later this year, Android 15 phone models could have a fuller AI experience than iPhone15, given Apple’s loyal customer base, we do not expect many ‘switchers’ from iPhone.Having said that, we think Google has a head start over Apple (and other hyperscale peers) in Google’s top-down approach to its own LLM, its own AI accelerators, fully owned data centers, cloud AI computing fully integrated into AI client device.

Google vs. MSFTIt would be interesting to watch how MSFT deals with the delay at NVDA’s Blackwell. We think investors do not appreciate the negative impact to MSFT long-term planning.Google on the other hand is master of its own AI universe. We think its plans to ramp up its next generation TPU6 silicon and its plans for AI infrastructure capacity expansion are on schedule.