> ## Content Index > Fetch the complete content index at: https://spyglass.org/llms.txt > Use this file to discover other available public pages before exploring further. # NVIDIA's "Delight" - URL: https://spyglass.org/nvidia-gpu-vs-google-tpu/ - Published: 2025-11-26T14:17:05.000Z - Updated: 2026-01-26T16:30:46.000Z - Description: Passive-aggressive tweets aside, NVIDIA has a point, but also a potential problem... - Author: M.G. Siegler - Tags: nvidia, jensen huang, google, gemini, google cloud, ai, meta, tech You know how I know that NVIDIA is concerned about Google's encroachment on their market? They won't shut up about how they're *not* concerned about it. As relayed by [Kif Leswing of *CNBC*](https://www.cnbc.com/2025/11/25/nvidia-says-its-gpus-are-a-generation-ahead-of-googles-ai-chips.html?ref=spyglass.org): > [Nvidia](https://www.cnbc.com/quotes/NVDA/?ref=spyglass.org) on Tuesday said its tech remains a generation ahead of the industry, in response to Wall Street’s concerns that the company’s dominance of AI infrastructure could be threatened by [Google’s](https://www.cnbc.com/quotes/GOOGL/?ref=spyglass.org) AI chips. > > “We’re delighted by Google’s success — they’ve made great advances in AI and we continue to supply to Google,” Nvidia said in a [post](https://x.com/nvidianewsroom/status/1993364210948936055?ref=spyglass.org) on X. “NVIDIA is a generation ahead of the industry — it’s the only platform that runs every AI model and does it everywhere computing is done.” I feel like a psychoanalyst could work wonders on that statement – including the fact that it's seemingly meant to be the royal "we", as it was [sent from the corporate account](https://x.com/nvidianewsroom/status/1993364210948936055?ref=spyglass.org), but it sure *feels* like it was written by or for one person who happens to be in charge of said company. Anyway, I don't know Sigmund Freud or Carl Jung, but I'm pretty sure my friend ChatGPT does, so I asked them to analyze the statement: > That has a passive-aggressive *flavor*, even though it's wrapped in corporate PR language. Here’s why: > > **💬 What the message says (on the surface)** > > • “We’re delighted by Google’s success.” > • “We continue to supply to Google.” > • “We’re the only platform that runs every AI model.” > > All of that is framed as positive, collaborative, upbeat. > > **😏 What the message *implies* (between the lines)** > > **• “Congrats to Google… but remember, they still rely on *us*.”** > **• “Google is succeeding thanks in part to our hardware.”** > **• “Whatever Google built, we’re still a generation ahead.”** > **• “They’re great… but we’re better, and essential.”** > > That contrast — *praising someone while subtly asserting superiority* — is a classic form of corporate passive-aggression. > > **Why it reads passive-aggressive** > > • The praise feels *performative*, not genuine. > • The second sentence immediately pivots to **asserting dominance** (“NVIDIA is a generation ahead…”). > • The phrasing implies: *“Let’s be clear: Google’s success doesn’t undermine us. In fact, it shows how central we are.”* Yeah, that's roughly my read. And the [read](https://x.com/ZuhaibHanfi/status/1993542367606198671?s=20&ref=spyglass.org) of [more](https://x.com/CaesarCapitalz/status/1993367530661200122?s=20&ref=spyglass.org) or [less](https://x.com/HimanshiTech/status/1993458435640508777?s=20&ref=spyglass.org) [everyone](https://x.com/imejasmin/status/1993533714543132722?s=20&ref=spyglass.org) on the [internet](https://x.com/lulumeservey/status/1993485593821823463?ref=spyglass.org). And thus marks [the *second day in a row*](https://spyglass.org/nvidia-evokes-the-e-word/) that NVIDIA has sort of stepped in it with an overly aggressive (even if passive) self-own. If the first two rules of Bubble Club are not to talk about Enron, [the third rule](https://x.com/mgsiegler/status/1993372731526590592?ref=spyglass.org) is probably not to so overtly betray your fears in public under the thin veneer of competitive praise. Look, we get it. After a stock run the likes of which has never been seen at this scale, culminating in not just NVIDIA becoming [the most valuable company in the world](https://spyglass.org/ai-divine-nvidia/), but being the first stock to surpass both a $4T *and a $5T market cap*, the past couple of weeks have been rough. But "rough" is also relative, *NVIDIA is still a $4.3T company*, and *still* the most valuable company in the world. But yes, [Apple and... *Google* are quickly catching up](https://spyglass.org/big-techs-wild-market-cap-ride/) as NVIDIA falls. It's a CEO's job to worry about such things, but it must be especially worrying when your best mechanism to combat this – and by far the hardest to pull off – [crushing your earnings expectations](https://spyglass.org/nvidia-big-tech-money-funnel/), didn't really work to combat the drop.[1](#one) And that's a problem because it suggests the market is shifting from exuberance – be it rational or irrational – to irrational fear. And if that takes hold... watch out below. [Not just NVIDIA, but everyone](https://spyglass.org/the-ai-bubble/). But *especially* NVIDIA. So that's *clearly* why we're getting a response to [the concerns by short-sellers](https://spyglass.org/punxsutawney-jensen/) and now to the market's reaction to [Google's recent success](https://spyglass.org/gemini-vs-chatgpt-product/). This is Jensen Huang trying to reassure everyone that everything is all good and quite rational. But *especially* NVIDIA. And he might not be wrong! Back to Leswing: > In its post, Nvidia said its chips are more flexible and powerful compared with [so-called ASIC](https://www.cnbc.com/2025/11/21/nvidia-gpus-google-tpus-aws-trainium-comparing-the-top-ai-chips.html?ref=spyglass.org#:~:text=Custom%20ASICs%2C%20or%20application%2Dspecific,companies'%20reliance%20on%20Nvidia%20GPUs.) chips — such as Google’s TPUs — which are designed for a single company or function. Nvidia’s latest generation of chips are known as Blackwell. > > “NVIDIA offers greater performance, versatility, and fungibility than ASICs,” Nvidia said in its post. There is certainly a world in which the flexibility that NVIDIA's GPUs offer over Google's TPUs ends up mattering. And perhaps *a lot* – especially if we're on the cusp of moving beyond, or even just *extending* from, being mainly focused on LLMs and instead are about to veer off into other types of AI models, [such as "World Models"](https://spyglass.org/llms-vs-world-models-ai/). (Though it's also clear that [Demis Hassibas is himself focused on such models](https://spyglass.org/signal-google-roars-back/#:~:text=%F0%9F%8C%8E%20World%20Models%2C%20So%20Hot%20Right%20Now) and as such, undoubtedly has Google's chips focused on how to train them as well.) Google's TPUs look great right now because they seem to be rivaling what NVIDIA's GPUs can do in terms of model outputs – based solely on the latest versions of Gemini. And even if NVIDIA is "a generation ahead", the TPUs are likely quite a bit more *efficient* by their nature of being purpose-built. Another fear (and eye-opening moment) in this news cycle for NVIDIA is probably the references to the notion that Google may have [cracked something new in pre-training](https://x.com/OriolVinyalsML/status/1990854455802343680?ref=spyglass.org) whereas the other models seem sort of "[stuck](https://spyglass.org/the-ai-wall/)" on their current trajectories in that regard. And that *could* imply [there's something about](https://spyglass.org/signal-google-roars-back/#:~:text=%F0%9F%92%B8%20Sam%20Altman%27s%20%27Rough%20Vibes%27%20Memo) the TPUs that helped with that breakthrough, whatever it may be. Or it may not imply that at all! [Nobody knows](https://www.youtube.com/watch?v=JYqfVE-fykk&ref=spyglass.org). Well, *someone* does. But they're not talking. So the rest of us are left guessing. (Or *are they talking* in the form of passive aggressive tweeting? Just asking...) NVIDIA, of course, has absolutely owned the pre-training portion of AI to date – so much so that the other would-be competitors are mainly focused on other parts of the build-out, such as inference. So much so that they have become a $5T company! And while Google is believed to have been using TPUs to train Gemini for at least a couple of generations now, the fact that they still are one of the largest buyers of NVIDIA chips suggests there's still *something* they felt the need to use GPUs for as well – as NVIDIA implies in their statement! Though that's a bit gray as well since a lot of Google's NVIDIA bill is undoubtedly to be able to host their chips on Google Cloud for *others* to use... > “We are experiencing accelerating demand for both our custom TPUs and Nvidia GPUs,” a Google spokesperson said in a statement. “We are committed to supporting both, as we have for years.” That's an interesting and seemingly very specific phrasing and framing of a PR statement in response to NVIDIA's own PR statement! Such wording *seems* to suggest that Google's need for NVIDIA's chips are not for their own purposes but rather for their customers (of Google Cloud). But again, it's sort of gray! Undoubtedly on purpose! Anyway, [the reports](https://www.theinformation.com/articles/google-encroaches-nvidias-turf-new-ai-chip-push?rc=lsmcir&ref=spyglass.org) that not only is Meta potentially partnering with Google to use TPUs, but is said to want *to acquire them to use them for training* of [new models](https://spyglass.org/meta-ai-world-models-lecun-agi/)... well, it certainly makes you at least think about NVIDIA's ongoing dominance there. > Unlike Nvidia, Google doesn’t sell its TPU chips to other companies, but it uses them for internal tasks and allows companies to rent them through Google Cloud. But they're [clearly](https://spyglass.org/signal-stock-performance-remedies/?ref=spyglass-signal-newsletter#:~:text=%F0%9F%8D%AA%20Google%20TPUs%20Move%20More%20Into%20NVIDIA%27s%20Lane) about to... And we're quickly going to learn and *hear* a lot about price, performance, power consumption, etc. Any of which *could* be bad news for NVIDIA. And any *combination* of which could be even worse news.[2](#two) For the first time, there could be real competition in the market. Imagine that. And you can probably just follow NVIDIA's *actions* here to get a sense of how they *actually* feel about Google entering the market. From [*The Information* report by Amir Efrati, Erin Woo, and Anissa Gardizy](https://www.theinformation.com/articles/google-encroaches-nvidias-turf-new-ai-chip-push?ref=spyglass.org&rc=lsmcir): > Whether Google succeeds in its TPU efforts, the mere specter of a powerful alternative to Nvidia may have already yielded benefits for big Nvidia customers like Anthropic and OpenAI that don’t want to rely on a single AI chip provider. > > Last month, following a Google deal to provide up to 1 million TPUs to Anthropic, Huang announced another deal to invest billions of dollars into Anthropic. He also got a commitment from the AI startup to use Nvidia GPUs. > > Similarly, after it became publicly known OpenAI was planning to rent TPUs from Google Cloud, Huang struck a tentative deal to invest up to $100 billion in the ChatGPT maker so it could develop its own data centers, and has discussed leasing out Nvidia GPUs to the company. Yeah, we're [totally about to see](https://spyglass.org/big-tech-ai-investments/) some sort of [massive](https://spyglass.org/nvidia-intends-to-invest-up-to-100b-in-openai-over-time/) NVIDIA/Meta deal, aren't we? At the same time, the fact that the entire AI chip market is still supply constrained suggests NVIDIA's business probably doesn't have too much to worry about; certainly not in the short term. And if Google's pre-training breakthrough isn't something specific to their TPUs, it also suggests that LLM scaling, as we've known it to date, will likely continue for a while yet. And that's good news for the status quo, which is naturally good news for NVIDIA. Even if Google is able to start selling TPUs to third-parties, it will undoubtedly take a long time to ramp such production – probably *years*.[3](#three) And in this market, no one can afford to wait, quite literally. But all of the above is nuanced and hard to um, grok – certainly for most mom & pop investors, but even some Wall Street analysts, it seems. So NVIDIA is going down their own defensive path here. Which is misguided, but you get why they're doing it. The market has some jitters and the company atop the market is doing their best to try to calm those to keep the good times rolling. So sure, NVIDIA is "delighted by Google's success" in many ways. But they're also less "delighted" in other ways, naturally. And that relationship is seemingly going to get more complicated going forward... [![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/2025/11/G6oLhWYbwAAKxEt.jpg)](https://x.com/andrew%5Fn%5Fcarr/status/1993418458877641208?s=20&ref=spyglass.org) credit: [@andrew\_n\_carr](https://x.com/andrew%5Fn%5Fcarr/status/1993418458877641208?s=20&ref=spyglass.org) **Disclosure: I worked at Google for 11 years as a partner at their venture fund, GV. Obviously, my thoughts are my own on these matters.* 👇 ****Previously, on** ***Spyglass** **...** [NVIDIA Evokes the ‘E’ WordBecause they invoked it! Multiple times! In a memo!![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/icon/Spyglass-Rings-----Multi-690.png)SpyglassM.G. Siegler![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/thumbnail/G6jU_qpWkAAJ2q5-1.jpg)](https://spyglass.org/nvidia-evokes-the-e-word/) [The Small NVIDIA ShortThe stakes keep rising for NVIDIA’s earnings as bears & bubbles loom…![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/icon/Spyglass-Rings-----Multi-691.png)SpyglassM.G. Siegler![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/thumbnail/3E09DDDE-B86F-491E-AB82-A6CAA7173E56-1-4.png)](https://spyglass.org/punxsutawney-jensen/) [“Hey, There’s a Bubble.”Yes, it’s a bubble. Yes, that may be a good thing. Yes, it will hurt.![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/icon/Spyglass-Rings-----Multi-692.png)SpyglassM.G. Siegler![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/thumbnail/ChatGPT-Image-Nov-10--2025--04_47_39-PM-4.png)](https://spyglass.org/ai-bubble/) [Meta AI vs. the World (Models)The writing was on the (Facebook) wall for Yann LeCun; now can he run an end run around Zuck to AGI?![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/icon/Spyglass-Rings-----Multi-694.png)SpyglassM.G. Siegler![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/thumbnail/ChatGPT-Image-Nov-12--2025--06_54_53-PM-5.png)](https://spyglass.org/meta-ai-world-models-lecun-agi/) [LLMs vs. the World (Models)A key schism in AI heading into 2026…![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/icon/Spyglass-Rings-----Multi-693.png)SpyglassM.G. Siegler![](https://storage.ghost.io/c/af/ca/afcaa655-46e2-45b8-889a-2881de5cce69/content/images/thumbnail/mgs22_an_asteroid_flying_across_the_sky_as_volcanoes_explode__fdd28b40-9b8f-49c2-ab1c-d80d6a8eb796_1-5.png)](https://spyglass.org/llms-vs-world-models-ai/) --- 1 One has to wonder if the true "break glass" mechanism NVIDIA tries to use here [involves China](https://www.bloomberg.com/news/articles/2025-11-21/trump-team-internally-floats-idea-of-selling-nvidia-h200-chips-to-china?ref=spyglass.org). If Jensen Huang can somehow convince the powers that be to let him sell there again it will give the stock bulls a new lease and outlook on life. But clearly he's having trouble convincing both the US *and* China to allow for this, despite [some extraordinary concessions](https://spyglass.org/the-art-of-the-quid-pro-quo/)... [↩](#) 2 My mind for some reason drifts to when Apple Silicon entered the CPU arena with a purpose-built chip to take on Intel. The incumbent kept touting top-end performance [without realizing](https://500ish.com/apples-embarrassment-of-intel-e2cf258ee320?ref=spyglass.org) that [the market had shifted](https://500ish.com/dial-m1-for-murder-f4e615a8a6bc?ref=spyglass.org) to prize efficiency more... I'm not saying this market will play out the same way, but it's certainly interesting to note. [↩](#) 3 And I'm not even getting into NVIDIA's software advantage here thanks to CUDA. Yes, many are trying to bridge that moat – including Google – but it's still a huge advantage their chips have that will take time to shift, if nothing else. [↩](#)