Rolf Bulk retweeted
Strong print by $AMD. DC CPU revenues set to double next year, driven by Venice on TSMC’s N2. Helios on track; we expect ~4GW of deployments in 2027, with revenues of $48B vs. ~$15.5B this year.
We continue to expect total revenues of >$100B in 2027, growing ~95% YoY, vs. consensus of ~$90.4B going into the print. This is despite the PC CPU market contracting a further ~10% next year.
Margin was the focal point given the topline beat, but we do expect some pressure there in 2027 as HBM costs increase in the mix. We expect AMD and peers will absorb part of that increase rather than fully pass it on to customers. Despite this, EPS set to more than double in 2027 to ~$19
Lot to unpack in the Anthropic S-1. Interesting data point on the share of revenue booked through CSPs: 48% of total. Our data shows Anthropic's penetration is ubiquitous across the different clouds, with a Net Score (a measurement of growth) of 82/100 in each of the 3 major CSPs, above OpenAI (53 to 56). OpenAI's early engagement with Microsoft hasn't translated into sustained higher penetration in Azure.
That is based on ETR's October TSIS survey of 1,684 enterprise IT decision makers, including 286 of the Fortune 500 and 416 of the Global 2000.
Anthropic cut prices to respond to increased competitive pressure, with a 40% lower cost per task for Opus 5.5 and 25-45% lower for Fable 5.1.
Our data shows cost is Claude's weak spot: Claude leads satisfaction at 4.5/5, but rates lowest on cost-effectiveness at 3.3, the widest gap of the top four AI tools we track. We know price elasticity (at least on paper) is high: 66% of users would move a production workload to open weights for a discount of 25% or less. Between proprietary models, the elasticity is likely even higher.
October ARR is the key metric to keep an eye on.
$AMAT increasing its stake in $Besi to 20%, up from 9% today.
Sensible investment. Besi has underperformed materially YTD as a result of the push-out of hybrid bonding adoption in HBM and recent newsflow around ASML entering the market for hybrid bonding (though that was known for quite some time).
However, hybrid bonding is still a no-brainer long-term for any chiplet-based architecture requiring high density interconnects, whether its GPUs, optics, HBM, smartphone SoCs, etc. In that market Besi is the clear leader: you don't book ~60-65% gross margins, higher than even ASML and KLAC, if you do not deliver consistent and superior value to your customers.
An acquisition may be challenging due to required regulatory approvals.
ir.appliedmaterials.com/news…
Rolf Bulk retweeted
“For YMTC, it’s the first time in over a decade where it’s able to command premium pricing for its products,” says @rolfbulk . “YMTC is now able to generate cash flow, shore up its balance sheet, invest in capacity, and all while maintaining very high margins. It’s a very good setup and a very opportune timing.”
Micron management is doing a good job by increasingly locking in SCAs at higher ASPs.
Eventually you'd expect the stock to re-rate given the higher share of locked in demand, but in the interim Micron management is probably happy to buy back stock at current multiples.
My initial read on why $MU isn't more green is that Q4 was still heavily price driven with DRAM revenue up 27% while bits only grew mid single digits and NAND up 42% with ~30% pricing so the market is still treating these earnings like a peak that fades once supply catches up.
What changed this quarter is management gave us a lot more visibility into why that may not happen anytime soon with 75%+ of next year’s output already committed, SCAs stretching into 2031, margins rising after Q1 and meaningful new cleanroom capacity still not arriving until late 2028.
If Micron can sustain something around $200 of forward EPS and market eventually decide that setup deserves even 10x instead of 6x then you're suddenly looking at $2,000 stock without needing an aggressive multiple.
Strong print by $MU on all fronts:
- Demand will continue to exceed industry supply, with shortages set to widen through CY27 and CY28. Even giving full credit to every fab currently planned by Micron, SK Hynix, Samsung and CXMT, we still see shortage persisting through 2030.
- Gross margin below expectation for 1FQ given compensation charges, but expanding through the rest of the year: ASPs are set to continue to increase.
- LTAs: now extending into 2031, new ones signed at higher prices.
- FCF this Q above $30B. Capital returns step up from December, with management committing to return 100% of excess cash.
We expect EPS above $200 in CY27. For CY28, consensus has ASPs declining materially and EPS declining to $170. We expect ASPs to remain elevated, driven by demand growth exceeding supply growth in both DRAM and NAND. Combined with the buyback, we see CY28 EPS closer to $250.
Happy to discuss the Micron print, broader memory sector, and Asia tech with Bloomberg this morning.
bloomberg.com/news/videos/20…
Great context by @PatrickMoorhead
I just chatted with $AMD's SVP of AI Vamsi Boppana. Here are some incremental adds to the previous post.
-Scope is AMD-wide. Not just physical AI. He called the team a nucleus to attract more talent. The insight is meant to come from model development at large, and the team was already heading into multimodal models.
-No frontier competition, by design. He said competing with customers at the frontier is "not the intent at all." He framed the non-language focus as a deliberate choice to stay complementary to customers' LLM work.
-Not an open-model push. He declined to frame it as an open-model strategy. AMD wants the capability for a world of many specialized models; one customer told him it needs 10,000. That's softer than Fei-Fei's public commitment to providing open models and platforms.
-Horizontal, with no lead segment. He named content and media, robotics and embedded, and data center prediction and reasoning, calling it "a horizontal models capability for the company." He didn't name a first product or a robotics customer.
-Built on AMD, deployed anywhere and on anything. Training will lean toward AMD, with more compute behind it. Deployment stays multi-platform because World Labs already has products running on other platforms. He gave no Instinct share.
-The roadmap link is in-house experimentation. His explanation was that he has often wanted people who could "model this and actually train something" to test hardware ideas. That's the mechanism; he gave no MI500, MI600 or Kria specifics and I didn't expect him, too.
-Talent and roadmap together drove the price. AMD saw World Labs' undisclosed roadmap and was "blown away." It already has a models team behind its Instella models and five NeurIPS papers, which this deal takes further.
-Commercial status quo for now. For Marble, the API and Autodesk, nothing changes near term, and how they evolve is TBD.
Rolf Bulk retweeted
Very disingenuous to suggest the industry is unilaterally suggesting AI needs to slow down.
Jensen and Dario are at opposite ends of the spectrum on this issue and many leading voices have disagreed with the idea of slowing down.
Safety must be prioritized, but the economic risks of slowing are palpable. Regulatory capture comes with its own risks. 😬
Rolf Bulk retweeted
A year ago, Anthropic trailed OpenAI by 10 pts with the Global 2000.
Today, it leads by 25. In the same survey, its midsize and small accounts fell 8 pts. The deepest budgets are deepening. The tail is pulling back. Preliminary October TSIS, still in the field.
$NVDA’s results and next-year guide were stellar, and its disclosures around the OpenAI leasing agreement and Amazon deal should ease concerns around ASICs vs. GPUs. This is not a zero-sum game: proprietary architectures will be deployed alongside GPUs at all major hyperscalers and AI labs.
Nvidia will need to cede some of its gross margin in an increasingly competitive environment. No surprise there, and when revenues are growing >70% to over $700B next year, not a major concern.
Happy to discuss the print in more detail with Bloomberg.
Have to admire $NVDA communication strategy today.
- Choosing to give early guidance for >70% revenue growth for 2027 will help drive the supply chain to accelerate multi-year fab build-outs.
- The guide also makes what is essentially a reset in gross margin expectations to low-70% from ~75% today palatable for investors.
- The disclosure of the 2 million GPU multi-year Amazon deal, days after Jalapeno, shows that even strong multi-generation ASIC programs such as Trainium can coexist with continued large-scale GPU deployments.
The guide of today should get Nvidia to ~$17 of EPS next year. Only 13x current $219 share price.
Very strong guide and definitely indicative we're moving more towards the high end of the $3-4T in annual DC Capex Nvidia expects for 2030.
70% growth means Nvidia is likely to exit next year at a $850-900B revenue run rate, more than 2x vs the current Q.
$NVDA expecting 70% growth in FY 2028 should massively deflate bearish sentiment around slowing growth. This is even with broader competition and why we see a bull case of over $14 Trillion in cumulative capex between now and 2030.
Rolf Bulk retweeted
Naura is bound to gain share, but in semicap share shifts are gradual. You cannot swap an Applied or Tokyo Electron tool with a Naura tool and expect identical performance. This is the crux of the Chinese semiconductor build-out; CXMT, Smic, YMTC will always prefer Western higher performance tools, as long as export restrictions allow for it.
Strong recovery rally in Korea today. Samsung up 24%, $SKHY 29% (!).
We'd expect volatility to remain high, but fundamentals are very strong: SK and Samsung are amongst the most profitable tech companies globally, memory remains in a structural shortage, and the stocks trade on 3-4x NTM earnings.
cnbc.com/2026/07/31/south-ko…
