Morgan Stanley highlights 4 risks to the FCC's consideration of Chinese transceiver ban:
1. Not enough capacity - 60-65% transceivers come from China2. China could potentially limit InP exports3. Lumentum and other laser makers would lose China demand 4. Marvell and Lumentum deal requires China's approval and that could face more headwinds$Marvell Tech(MRVL.US) $Coherent Corp.(COHR.US) $Lumentum(LITE.US)
Equity research
Articles
Mizuho highlighted 5 takeaways from their dinner with SemiAnalysis:
> ASICs closing the gap with GPUs.> Agentic-AI CPU demand headwinds with token consumption potentially slowing (e.g. Muse) as SA sees 50-70% 2027E CPU growth challenged.> Neocloud headwinds into 2H27E as SpaceX adds~13GW in 2027E.> SA seeing a significant HBM portion on 4-Hi, while 12-20Hi potentially pushed further out or canceled. Alsonoting slower NAND inference rack demand, potentially lowering the need for SLC/HBF.> China InP oversupply headwinds into 2027E.$AXT(AXTI.US) $SpaceX(SPCX.US) $Nebius(NBIS.US) $Coreweave(CRWV.US) $Broadcom(AVGO.US) $Alphabet(GOOGL.US) $AMD(AMD.US)Current tape:
> Capex peak in 2027 fears > Token spend flat > OpenAI ARR possibly flat> Semis margin contraction in 2027/2028> FED rate hikes in December possibly tightening financing even more I've been in a lot of cash now and probably the best idea to keep that up. This is a falling knife market.Barclays: AI lab revenue is estimated to grow to $376 billion in 2027. Relative to 2024 levels, this trajectory implies revenue growth of 271% in 2025, 427% in 2026, and 174% in 2027.
$Alphabet(GOOGL.US) $Meta Platforms(META.US) $Microsoft(MSFT.US) $Anthropic(ANT.US) $Oracle(ORCL.US) $Amazon(AMZN.US)SemiAnalysis: CapEx
"We revise down our forecast of AI chip/server shipments, spend, and power demand across our forecast horizon, with forecasts more significantly impacted from 2028 to 2030."Reasons for the cut include:> Slowing Q3 net-new ARR growth at frontier AI labs requiring a reset to ensure token revenues sustainably cover compute costs.> Tightened debt and equity capital markets reducing financing availability for non-hyperscaler-backed datacenter developments.> US midterm election results delaying 5.8 GW of planned US datacenter supply from late 2027 and early 2028.> Normalizing HBM supply and demand driving down average selling prices and nominal accelerator capex.> Severe power constraints forcing vendors to moderate maximum chip power targets across future product roadmaps.> Shifting performance-per-TCO and financing-backed preferences at frontier labs away from marginal Nvidia platforms toward custom ASICs and alternative hardware.> Heightened financing, execution, and export control risks surrounding long-tail neocloud customers with high Nvidia exposure.$Alphabet(GOOGL.US) $Broadcom(AVGO.US) $NVIDIA(NVDA.US) $Meta Platforms(META.US) $Microsoft(MSFT.US) $Amazon(AMZN.US)Bloomberg: Nvidis Faced Pressure to Boost Its AI Backstop Before Huang Exit
"After months with no deals arising from a $500 billion partnership with six of the world's largest asset managers, Nvidia in the past month has faced calls from some customers and Wall Street partners to boost the size of its financing backstop, according to people familiar with the matter, who asked not to be named discussing private negotiations."$NVIDIA(NVDA.US)Morgan Stanley: Google’s TPUs are expected to account for 35% of global demand, slightly below NVIDIA’s 36%, with the two companies combined reaching 71%. AMD is projected to hold 16%, Amazon AWS 5%, customers of Global Unichip Corporation approximately 3%, and Microsoft and Meta each around 2%. In terms of chip types, GPUs are expected to constitute 52% of HBM demand, while custom AI chips (ASICs) will account for 47%, indicating a near-even split.
$Micron Tech(MU.US) $DRAM $EWY $SK Hynix(SKHY.US)Yuanta Securities: Accelerator Market
> Nvidia continues to hold leadership in shipment volume, with nearly 15 million units in 2027F. Google a close 2nd. > Nvidia's accelerator market share in terms of shipment volume expected to be 54% in 2026F, before dropping 41% in 2027F as non-Nvidia chips ramp in higher volume. $NVIDIA(NVDA.US) $Amazon(AMZN.US) $Alphabet(GOOGL.US) $Meta Platforms(META.US) $Microsoft(MSFT.US) $AMD(AMD.US)Menlo Ventures: Enterprise AI
"For every $1 the industry spends building AI models, enterprises spend about $1.70 putting them to work.""We estimate that Anthropic and OpenAI alone account for 25%+ of AWS, Azure, and Google Cloud revenue, with much of that spend going to training. That investment only pays off if enterprises keep putting models to work, which agents are now driving."$Amazon(AMZN.US) $Alphabet(GOOGL.US) $Microsoft(MSFT.US) $NVIDIA(NVDA.US) $AMD(AMD.US) $Meta Platforms(META.US)Deutsche Bank sees a potential slowdown in the overall expansion of computing power in Texas as a benefit for SpaceX.
The report notes that SpaceX has demonstrated the ability to build computing capacity faster and at lower cost than its competitors. Its new neocloud contracts with customers include a 90-day cancellation clause, allowing the company to flexibly capture upside when spot market prices rise sharply.Deutsche Bank believes that if the industry's overall computing power supply is constrained, SpaceX has the ability to significantly raise its prices over the next few years, thereby boosting revenue and EBITDA while reducing capital expenditures.$SpaceX(SPCX.US) $Nebius(NBIS.US) $IREN(IREN.US) $Coreweave(CRWV.US) $Alphabet(GOOGL.US) $Oracle(ORCL.US)NYT: Commerce Dept. Chip Inquiry Was Halted After Nvidia Intervention
A representative for Megaspeed said the company was conducting an internal investigation into allegations of chip diversion and had repeatedly offered to cooperate with the U.S. government but that those offers had gone unanswered.The United States began investigating Megaspeed and Ms. Huang in 2024, after the company started importing billions of dollars’ worth of advanced Nvidia A.I. equipment into Malaysia. Megaspeed had been planning and constructing data centers around Southeast Asia to serve Chinese clients, including Alibaba, the Chinese tech giant. While that behavior is legal, U.S. officials were concerned that some of the Nvidia chips the company had bought were illegally routed into China.In August 2025, officials at the Commerce Department’s Bureau of Industry and Security who were pursuing the case received an email from Mr. Teter, the general counsel of Nvidia, people familiar with the incident said. In a message to the agents, Mr. Teter said that Mr. Kessler had given him permission to reach out about the case.A Commerce Department official said that Nvidia had contacted the employees directly to request a meeting with one of them. When the employees declined the meeting, Mr. Teter responded to the employees by saying he was unsure what the issue was because Mr. Kessler had said Nvidia could communicate with the bureau’s employees.The employees forwarded the email chain to Mr. Kessler. Hours later, Mr. Kessler responded, saying they should continue to handle the case strictly by the book, the official said.Mr. Kessler did not order his employees to divulge information to Nvidia, and he did not weigh in with the agents on what meetings to take or what information to divulge, that official said.$NVIDIA(NVDA.US)Expert Takeaways: Amazon AWS
> Incremental physical AI capacity stands at 55% AWS Trainium versus 45% NVIDIA GPUs.> Projected net-add AI capacity of 7.3 GW breaks down physically to 60% Trainium and 40% NVIDIA hardware, translating to a roughly 50/50 capital-expenditure split due to NVIDIA's ~1.5x higher per-GW capital cost.> Current Trainium capacity is split 60% inference and 40% training, with inference share expected to rise to 70% in 2027 as legacy Trainium 2 clusters are repurposed and pre-training jobs migrate to Trainium 4 clusters toward late-2027.$Amazon(AMZN.US) $NVIDIA(NVDA.US)If anyone is curious about the pace of deterioration AI infrastructure names are seeing from their 52 week high. Here is a table of companies in the WFE, accelerators, neoclouds, optical, and memory market.
Best performer? Nvidia, -25% Worst performer? Applied Optoelectronics, -80% This is a non-exhaustive list. $Micron Tech(MU.US) $NVIDIA(NVDA.US) $Lumentum(LITE.US) $AMD(AMD.US) $Dell Tech(DELL.US) $Broadcom(AVGO.US) $Lam Research(LRCX.US)Jeongdong Choe, Senior Vice President of the Semiconductor Division at TechInsights, expects the memory boom to last for 3-4 more years.
> Supply currently meets only about 85% of demand. > DRAM supply is short by approximately 10–12%, with NAND facing a similar deficit.> SVP Choe anticipates that supply levels will rise to 90–95% of demand by next year. Given that a balanced market typically requires supply to exceed demand by 2–3%, a seller's market is likely to persist.> A potential supply-driven downturn might re-emerge in 5–6 years, but that remains a long-term risk not currently weighing on short-term outlooks.> Supply at 100% of demand is still technically tight; true equilibrium requires supply at 102–103%.> In the second half of next year, as HBM4 adoption ramps, Samsung's market share could rise to 25–30%.> In DRAM, the process gap between Korean and US producers has largely closed.$Micron Tech(MU.US) $DRAM $EWY $SK Hynix(SKHY.US) $Sandisk(SNDK.US)Rosenblatt: "We believe each $1mn of InP substrates can support ~$30mn of transceiver sales, so the $194.4mn 2027 contract could underpin ~$5.8bn of annual transceiver revenue, or roughly 12mn modules at the ~$500 blended ASP implied by AAOI's 2027 revenue and capacity targets. Over the full five-year agreement, that implies ~$29bn of transceiver revenue at constant pricing."
$Applied Optoelectronics(AAOI.US)Nomura: laser chip shipment volume is forecasted to grow from 480 million units in 2026 to 966 million units in 2027, and reach over 1.5 billion units by 2028. CW laser experiences the largest expansion, growing by 402 million units (from 171 million in 2026F to 573 million in 2028F).
$Coherent Corp.(COHR.US) $Lumentum(LITE.US) $Applied Optoelectronics(AAOI.US) $Broadcom(AVGO.US)SemiAnalysis sees 30.3 GW of planned 2028 capacity as risk to moratorium if the candidate proposing a datacenter pause won in every state where one is running.
"If the candidate proposing a datacenter pause won in every state where one is running, those states would together contain roughly 30 GW of planned capacity scheduled for 2028. How much actually slips depends on the permits each project still needs and how long the pause lasts."$Meta Platforms(META.US) $Alphabet(GOOGL.US) $Amazon(AMZN.US) $Microsoft(MSFT.US) $Oracle(ORCL.US) $SpaceX(SPCX.US)LightCounting: Although we expect scale-up switching will grow fastest, Ethernet will still deliver strong sustained growth with a projected 28% CAGR. All told, 2026 data-center switch sales are projected to grow 162% year-over-year in aggregate, and our forecast shows a 5-year CAGR of 36%. By 2031, Ethernet, InfiniBand, and scale-up switch ASICs should generate a combined $147 billion in sales.
$NVDDA $Broadcom(AVGO.US) $Alphabet(GOOGL.US)