Oguz Erkan
Investor, Founder of Capitalist Letters
Nvidia's biggest customers scaling their own chips threatens its long-term weight in a portfolio. Microsoft, Meta, Amazon and Google are building proprietary silicon; as AI workloads shift from training to inference, the shift toward ASICs should accelerate, with ASIC volumes expected to surpass GPUs by 2030. That is why Nvidia should not be overweighted for the long term. x.com
Sravan Kundojjala
Semiconductor Industry Analyst at SemiAnalysis
MediaTek 4Q26 ... 2027 ASIC is now at the high end of the $16B target (20% of the AI ASIC TAM) x.com
AlvaSupportsThe outlook for 2027 ASIC revenue near $16 billion strengthens the case that suppliers of custom accelerators are scaling into substantial competitors to Nvidia.
The Transcript
Weekly Earnings-Call Quotes Newsletter
$GOOGL CFO: ... In Google Cloud, we expect TPU system sales to continue to scale in Q4, followed by a meaningful revenue ramp in 2027 x.com
AlvaSupportsThe expected Q4 expansion and meaningful 2027 TPU sales ramp strengthen the case that Google's proprietary chips are becoming a larger external alternative to Nvidia.James A. Engle
podcast
We did not fully understand the lock in that they had with CUDA software system. And we finally came to the belief that that was a powerful thing. traffic.libsyn.com
AlvaChallengesThe reassessment of CUDA's lock-in strengthens the case that software switching costs can protect Nvidia against customers' proprietary chips.
Reuters
Global News Agency
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 shipments and expected Q4 accelerator revenue of $2.5 billion indicate commercial scale for another supplier targeting hyperscalers' custom AI chips, strengthening the competitive threat to Nvidia.
Amazon.com, Inc.
E-Commerce, Cloud Computing (AWS) and Digital Advertising Company
Enterprises are still very early in using inference of scale, and yet inference already accounts for well over half of AI revenue and growing.
AlvaSupportsInference's growing majority of Amazon's AI revenue strengthens the thesis's premise that inference is becoming the dominant commercial AI workload.
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." x.com
AlvaSupportsTrainium revenue commitments growing more than twentyfold year over year strengthen the evidence that customers are committing to Amazon's proprietary AI accelerators.Fireside Alpha
Business, Technology and Investing Interview Digest
Amazon CEO Andy Jassy on the chip business: ... “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
AlvaSupportsPlanned Trainium availability outside AWS in late 2027 would extend Amazon's proprietary accelerators into workloads beyond its own cloud, broadening competition with Nvidia.
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 of nearly two ASICs per GPU at AI labs by 2028 supports—and brings forward—the thesis's expectation that ASIC volumes will overtake GPUs by 2030.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. x.com
AlvaSupportsOpenAI's own confirmation that it has made an efficient chip strengthens the case that major AI buyers can develop proprietary alternatives to Nvidia.Fireside Alpha
Business, Technology and Investing Interview Digest
Oracle co-CEO explains why he doesn’t see GPUs outrunning demand anytime in the next three to five years: x.com
AlvaChallengesExpected GPU demand through roughly 2029–2031 provides a counterweight to the thesis that inference-driven ASIC adoption will diminish Nvidia's long-term role.
Zephyr
AI & Semiconductor Analyst at Citrini Research
@P_Remarks @BearForce_Won Well, OAI/Meta/google have also dedicated similar amounts of compute to R&D (60%-75%) ... Even the OS labs like DeepSeek have 70% of their fleet dedicated to training x.com
AlvaChallengesThe reported 60%–75% R&D compute allocations at OpenAI, Meta and Google, and 70% training allocation at DeepSeek, weaken the premise that inference is already displacing training as the dominant workload.
Dow Jones
Real-Time Financial News Wire
Nvidia's graphic processing units, or GPUs, dominated the initial push to train AI models. But companies like FuriosaAI are betting that their energy-efficient specialty chips could be more competitive for the next stage of AI inferencing, or using AI models after they're trained. tradingview.com
AlvaSupportsSpecialty-chip makers targeting inference support the thesis that AI workloads could shift away from Nvidia GPUs, though the item reports a bet on future competitiveness rather than demonstrated displacement.
Satya Nadella
Chairman and CEO of Microsoft
We continue to scale our fleet with the latest innovations from NVIDIA, AMD alongside our own Maia systems. ... 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.
AlvaMixedMicrosoft's Maia chips show meaningful efficiency gains that could support shifting some AI workloads away from GPUs, while the company is still scaling its fleet with NVIDIA and AMD systems.
Alphabet Inc.
Google's AI Prototyping Platform
Tippu 80 and eight AI offer up to 2.7 times better training performance and 80% better inference price performance respectively, versus our previous generation of Tpa's.
AlvaSupportsGoogle's claimed 80% inference price-performance improvement over its previous TPU generation strengthens the economics of its proprietary chips as inference workloads expand.The Transcript
Weekly Earnings-Call Quotes Newsletter
Google Cloud CEO Thomas Kurian: ... "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
AlvaSupportsGoogle's TPU inventory is backed by signed customer contracts, strengthening confidence that its external chip expansion reflects committed demand.Dow Jones
Real-Time Financial News Wire
Google Cloud CEO: Nearly 100 Customers Currently Using TPUs, With Strong Demand From 'Many Others' tradingview.com
AlvaSupportsNearly 100 customers already using TPUs gives concrete scale to external adoption of Google's proprietary silicon, strengthening the case that it can compete for workloads beyond Google's own.Reuters
Global News Agency
ALPHABET INC.: SEEING STRONG DEMAND FOR OUR TPU SYSTEMS ACROSS A WIDE RANGE OF CUSTOMERS tradingview.com
AlvaSupportsStrong TPU demand across a wide customer base supports broader adoption of Google's proprietary silicon as an alternative to Nvidia hardware.
Oracle Corporation
Oracle Enterprise Cloud Computing Platform
Do I think that every customer will suddenly want to buy Tpa's? No, that's okay. ... We have a lot of different options ... demand for all of it is far exceeding the ability to go out and supply it, at least for what I see for the next several years.
AlvaChallengesOracle expects demand across its accelerator options to exceed supply for several years, suggesting TPU adoption can coexist with sustained demand for competing platforms rather than quickly displacing them.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 contributing meaningfully to cloud revenue growth provide company-reported evidence that Google's proprietary chips are becoming a material commercial business.
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. x.com
AlvaMixedThe reported tripling of Anthropic's 2028 AMD orders would expand competition for Nvidia, while showing a major buyer increasing GPU deployments alongside its custom-chip commitments.