Master of Crypto
Crypto Market Intelligence Publisher
Nvidia's biggest long-term advantage faces its toughest test as AI companies build their own chips. DeepSeek is reportedly developing its own inference chip while raising $7B at a $52-59B valuation, and inference is one of the fastest-growing parts of AI. OpenAI, Google, Meta and Amazon are also investing in their own chips. Export controls have already cut Nvidia's China business from 20-25% of revenue to about 9%, with Huawei filling much of the gap. x.com
Nvidia's stock fell despite a strong quarter because the good news was already priced in. Revenue was $81.6 billion, guidance was $89-92B, it announced an $80B buyback and raised the dividend to $0.25, but almost everyone was already buying and few new buyers were left. The AI business is still very strong; the problem is that the stock ran up too much before earnings. x.com
Ranjan Roy
VP, Industry Lead for Retail at WRITER, Co-author of Margins
This doesn't help the overall business story that frontier models at high margin are going to be what differentiates them. ... In the enterprise side of it, the number of people talking about or working on routing models across whatever the lowest cost best model is for the task is definitely increasing. ... It certainly presents a threat, but it hasn't affected them yet, it looks like. podcasts.apple.com
AlvaSupportsIncreasing enterprise use of routing to the lowest-cost model makes inference more price competitive and raises the risk that Nvidia’s platform advantage will matter less, although the source says the threat has not yet affected the businesses discussed.
Reuters
Global News Agency
Chinese chip designer MetaX posted a third-quarter net profit as revenue more than quadrupled from a year earlier, helped by higher sales of graphics processing units, or GPUs, used in artificial-intelligence computing. ... China's domestic AI-chip makers are seeking to expand sales as US restrictions on exports of cutting-edge GPUs have encouraged Chinese companies to look for local alternatives, although demand for AI computing chips still far exceeds domestic supply. tradingview.com
AlvaMixedThis supports the thesis that export controls are accelerating Chinese alternatives to Nvidia, but the reported supply shortfall shows domestic competition is still constrained.
Sravan Kundojjala
Semiconductor Industry Analyst at SemiAnalysis
2027 ASIC is now at the high end of the $16B target (20% of the AI ASIC TAM) x.com
AlvaSupportsThe reported 2027 ASIC outlook at the high end of a $16 billion target points to substantial growth in custom-chip supply, strengthening the competitive challenge to Nvidia.Reuters
Global News Agency
Rick Tsai, CEO of Taiwan's largest chip designer MediaTek ... Tsai said demand for its first AI accelerator remains strong, with initial shipments starting last month to generate $2.5 billion of revenue in Q4 tradingview.com
AlvaSupportsInitial accelerator shipments expected to generate $2.5 billion in quarterly revenue show another aspiring custom-chip supplier reaching commercial scale, increasing competitive pressure on Nvidia.
Amazon.com, Inc.
E-Commerce, Cloud Computing (AWS) and Digital Advertising Company
Commitments for Gravitant and Trainium have each increased more than 20 times year over year. ... Trainium with the meaningful advantage it has in price performance versus alternatives are very compelling to people. ... We believe we're going to have a very significant business selling chips externally, and we're going to start early 2027.
AlvaSupportsThe call transcript puts Amazon’s planned external chip sales in early 2027, earlier than the late-2027 timing previously reported, potentially accelerating competition for Nvidia’s customers.
Fireside Alpha
Business, Technology and Investing Interview Digest
Amazon CEO Andy Jassy on the chip business: ... “Finally, our chips business continues to grow triple-digit percentages year over year, and revenue commitments for Graviton and Trainium have each increased more than 20 times year over year." ... “With Trainium demand so strong, customers continue encouraging us to make it available outside of AWS. We intend to do so late in 2027.” x.com
AlvaSupportsMore than twentyfold growth in Trainium revenue commitments and plans to offer it outside AWS in late 2027 strengthen the case that Amazon’s custom silicon can compete for a wider share of Nvidia’s market.John Coogan
podcast
I think we saw that from the AMD coverage we did with Lisa Su, where Anthropic Vibe coded their AMD implementation pretty quickly. So, good for anyone who has bet on oligopoly and silicon, rougher if you were like, this will not be a car like market, but still very, very early, still a lot of dependence on NVIDIA, obviously. podcasts.apple.com
AlvaMixedCoogan says Anthropic quickly implemented AMD support using AI-assisted coding, suggesting lower switching costs, while continued Nvidia dependence tempers the near-term displacement threat.
Damnang
Semiconductor Engineer in Silicon Valley, AI Chip Investment Research
AI labs will increasingly shift from GPUs to ASICs (XPUs), with the ASIC to GPU ratio rising from roughly 1:1 in 2026 to nearly 2:1 in 2028. x.com
AlvaSupportsThe forecast shift from roughly equal ASIC and GPU use in 2026 to nearly twice as many ASICs in 2028 strengthens the case that custom chips will take a larger share of AI labs’ compute.Reuters
Global News Agency
Alphabet is making billions of dollars selling homegrown AI chips it once reserved for internal use, but building inventory to meet surging demand is draining cash as the tech giant spends heavily on data centres. ... Sales of TPU systems, which bundle the chips with networking, storage and software, are expected to rise further in the current quarter and ramp up significantly in 2027, Google Cloud chief Thomas Kurian said. tradingview.com
AlvaMixedGrowing TPU sales and a planned 2027 ramp strengthen the competitive challenge to Nvidia, while upfront inventory spending is draining cash as Google scales the business.Fireside Alpha
Business, Technology and Investing Interview Digest
Sam Altman says ... We made a chip that is an incredible performer at units of intelligence per watt. We are learning how to build our own data centers. We have a robotics program that we’re going to point at making lots of data centers. x.com
AlvaSupportsOpenAI’s CEO confirms it has made its own chip and is building data-center expertise, strengthening the evidence that a major Nvidia customer is developing its own compute infrastructure.
Joey Brookhart
Tokenomics team co-lead at SemiAnalysis
Really token throughput, and they have a very diverse chip fleet. So between Trainiums, TPUs and GPUs, across all of those we hear a ton of token efficiencies. ... I think token throughput, for example, on Trainium chips are up five to 10 X year over year right now. podcasts.apple.com
AlvaSupportsBrookhart says Anthropic’s diverse Trainium, TPU and GPU fleet is producing major token-efficiency gains, including 5 to 10 times higher Trainium token throughput year over year, strengthening the case that alternatives can improve faster than Nvidia’s platform.
Dow Jones
Real-Time Financial News Wire
"Taking an advanced AI chip from design to volume production is an exceptionally difficult undertaking," said John Yetimoglu, chief investment officer at Infinitum. Few companies have demonstrated the ability to do that, but FuriosaAI has secured production deployments with leading enterprises such as Samsung SDS and LG Group, he said, backing Infinitum's conviction in FuriosaAI's ability to scale globally. tradingview.com
AlvaSupportsReported production deployments at Samsung SDS and LG Group show a specialized inference-chip rival gaining enterprise adoption, strengthening the case for alternatives to Nvidia.
Satya Nadella
Chairman and CEO of Microsoft
Maia 200 is already delivering up to 1.7 times more tokens per watt across the AI experiences in teams and Copilot and Maia 300 will drive further gains, delivering 40% more tokens per watt compared to the leading GPU based systems.
AlvaSupportsMicrosoft reports efficiency gains from Maia 200 in Teams and Copilot and expects Maia 300 to deliver 40% more tokens per watt than leading GPU systems, strengthening the competitive case for its in-house inference chips.
The Transcript
Weekly Earnings-Call Quotes Newsletter
Google Cloud CEO Thomas Kurian: Google is building TPU system inventory against committed customer orders, ... "When not mentioned that we procure components to build inventory, we only do that after we have signed contracts. So all of the inventory that for TPU systems are associated with committed contracts." x.com
AlvaSupportsSigned customer contracts back Google’s TPU inventory build, strengthening the evidence that its alternative to Nvidia has committed commercial demand.Reuters
Global News Agency
Google Cloud revenue surged 103% yr/yr, driven by strong growth in enterprise AI infrastructure and solutions ... Alphabet said sales of TPU systems contributed meaningfully to Google Cloud revenue growth tradingview.com
AlvaSupportsTPU system sales are now contributing meaningfully to Google Cloud growth, strengthening the case that Google’s custom chips are becoming a substantial commercial alternative to Nvidia.
Jukan
Semiconductor & AI Infrastructure Research at Citrini
Aletheia: Anthropic has increased its AMD GPU orders for 2028 to 3 GW—three times its original deployment target of 1 GW. ... Aletheia: OpenAI has increased its total planned AMD GPU deployments over 2026–2028 from 6 GW to 8–9 GW. x.com
AlvaSupportsReported increases in Anthropic’s and OpenAI’s AMD deployment plans point to growing multi-gigawatt demand for alternatives to Nvidia.
Bloomberg
Business and Financial Information and News Platform
AI chipmaker Moore Threads said sales grew 1,774% in the three months through September, buoyed by accelerating domestic demand for artificial intelligence hardware x.com
AlvaSupportsRapid sales growth at a Chinese AI-chip supplier strengthens the case that domestic alternatives to Nvidia are gaining commercial traction.Reuters
Global News Agency
It expects to begin limited production over the next two months, one person said. Yields, or the share of chips coming off a production line that meet specifications, are only about 20% to 30%, compared with yields of around 70% at South Korea's SK Hynix , the market leader, they said. tradingview.com
AlvaMixedCXMT’s planned HBM3E production could supply Chinese AI-chip designers, but reported yields of 20% to 30% versus about 70% at SK Hynix constrain the scale and cost of domestic alternatives.Reuters
Global News Agency
OPENAI PLANS TO BUY GOOGLE’S AI CHIPS tradingview.com
AlvaSupportsThe Information reports OpenAI plans to buy Google AI chips, indicating a leading model developer may shift some compute demand to a non-Nvidia supplier.
SemiAnalysis
Independent Semiconductor and AI Research Firm
“They have TPU, Trainium, NVIDIA GPU and AMD GPU all online. New hardware is always more performant, but you need the software to use it. x.com
AlvaMixedAnthropic reportedly has TPU, Trainium and AMD GPUs online alongside Nvidia GPUs, showing that non-Nvidia hardware is in use while Nvidia remains part of its compute fleet.