Ren
Head of AI, Ex-Product Manager, AI Supply-Chain Bottlenecks
Nvidia is winning fast inference: OpenAI shipped its flagship speed tier, GPT-6 Astra Ultrafast, on Nvidia GPUs, gaining a second source and eroding the exclusivity Cerebras was priced on, though Cerebras is still faster and capacity likely played a role. Operators keep extending GPU lives, Amazon to 5-6 years, Microsoft and Google to 6, Meta to 5.5, because old chips keep earning on inference. More lifetime revenue per chip justifies customers' next Nvidia order. x.com x.com x.com x.com
NVIDIA's moat widens as OpenAI's new GPT-6.1 Sol Ultrafast tier reportedly runs on its GPUs at low batch size. That shows GPUs can serve the speed tier too, weakening the case that fast inference needs wafer-scale chips or custom ASICs, and low batch means more GPUs per token, so growing speed tiers grow the GPU bill. Cerebras ran the previous version at up to 750 tokens per second versus about 300 on Nvidia; which hardware powers Ultrafast for GPT-6 Astra is the next tell. x.com
Nvidia's CUDA retains developers, putting it among 12 stocks that could dominate the decade ahead. Nebius's GPU rental price rise signals demand; Fabrinet and Credo supply AI connectivity. DraftKings has betting scale; Microsoft has Azure and an OpenAI partnership; Figma anchors design work; Shopify powers stores. TeraWulf pivots power sites to AI leases; Constellation, the largest US nuclear operator, signs big-tech power deals; Meta spends on AI; Uber has autonomous-vehicle partnerships. x.com … Show more
AI software and hardware advances make GPUs and power more valuable, creating huge long-term tailwinds for compute providers and AI model and app businesses. Each new software generation increases a GPU's value; each hardware generation increases the value of a gigawatt of power. That favors Nvidia, Broadcom, Intel and AppLovin. x.com
NVIDIA's GPU value pool faces architectural competition and shorter useful lives than ten-year assumptions imply. Etched and other accelerator designs may challenge performance within five years, while power limits favor relentless gains in performance per watt; past GPU generations improved more than 2-3x, and industry experts reportedly used four-year lives before GPT. The range of outcomes remains wide. x.com x.com x.com
More AI tokens mean more compute demand for Nvidia even as pricing power starts to slip. OpenRouter data via the FT show weekly token use up ~250-fold since January 2025; the FT links the surge to agents' long call chains. Bloomberg's breakdown of a 1 GW AI campus that could cost roughly $30B puts well over half the bill in chips and servers, most landing at Nvidia as GPU racks. Servers wear out in 5-6 years, making AI capex a recurring chip refresh. x.com x.com x.com x.com x.com x.com x.com
Nvidia GPUs may retain high residual values while memory, power, data centers and chips remain scarce. Ornn Data's chart shows B200 rentals in October still more than twice their January level, with older H200 rentals higher still. This may last longer than expected, but normal depreciation returns once newer generations ramp and supply catches up. x.com x.com
NVIDIA faces rival chips from every Mag 7 customer, which has an incentive to cut GPU spending and improve margins on its AI buildout. If they can make a GPU 85% as good for 50% less, it is worth using instead of NVIDIA's; the shift will take time. NVIDIA makes a GPU for $5,000 and sells it to those customers for $50,000, leaving them an incentive to save 90% by building their own. x.com x.com x.com x.com
Older GPUs retaining value after years of rental income is a positive for IREN's GPU-cloud economics. Per @wallstengine, an A100 system with 9,600 GPUs and servers is listed for $59.7 million, about $6,200 per GPU. Against an estimated original rack cost of $25,000 per GPU, that suggests a 25% residual after six to seven years; the GPUs also rented for $2-10 an hour for six years. The residual is based on an asking price, not a sale. x.com x.com x.com
Nvidia's system components can reduce the idle time that erodes returns on expensive AI chips. CPUs, accelerator interconnects and server networking help operators such as Nebius and IREN keep machines doing useful work; solving the next bottleneck could win Nvidia more of the data-center equipment budget even when a rival chip looks appealing alone. Buyers must compare the cost of completed work across the whole system. x.com
Nvidia's case is a lot weaker than two years ago, because large providers such as OpenAI and Meta are likely to design custom chips tuned to their own models. Nvidia wins only if workloads and models stay diverse. The claim that frontier labs are decoupling from Nvidia is doubtful, since OpenAI's recent efficiency gains likely came from Blackwell, but it holds for Google and Amazon. Nvidia also needs to own the open-source ecosystem, which is contradictory by definition. x.com x.com
Nvidia retains pricing power if its full system keeps reducing customers' cost per effective AI task. Strong AI demand alone does not ensure profits across the chain: rising chip, memory and optical costs burden buyers, while tighter financing and slow collections threaten operators. Large customers developing their own chips could pressure Nvidia's margins even as the market grows. x.com
