Nvidia's Rocket Ship ??- What did we just Witness?
Michael Spencer
A.I. Writer, researcher and curator - full-time Newsletter publication manager.
Hey Everyone,
Today Nvidia is a 1.942 Trillion company! What did we just witness?
We're rarely if ever witnessed the rise of a company like we have with Nvidia and the gigantic orders for H100 AI chips it received in 2023.
?? From our sponsor: ??
Monetize your AI Assistants
Nvidia is today in 2024 what Cisco was in 2000. Cisco was valued at a multiple of more than 150 times forward earnings in March 2000, when the stock hit its dot-com era peak.
As of today, Nvidia as a company is worth more than Amazon and Google, for the first time in a long time. First time since 2002 Nvidia is worth more than Amazon, for example.
The last few months of Nvidia, being a monopoly in AI chips defines the Generative AI hype movement. Here is the 5-year chart:
Even in a global economy where China, Japan and Germany are in recession or teetering on the brink of it, America appears red hot with technologies like OpenAI’s Sora that are striking fear in the hearts of Hollywood and the Chinese. None of which would be possible without Nvidia’s H100 AI chips.
Nvidia has become a meme-hyperscale stock, and you could make the argument its eating the rest of BigTech.
Nvidia’s event GTC on March 18th, 2024 will be key to understanding its future with the B100 AI chip. Nvidia said that it expected to be supply constrained with its next-generation Blackwell-based GPU products.
It’s doubtful Big tech companies such as Amazon, Microsoft, Google, and Meta are able to increasingly effectively reduce their dependencies on Nvidia, as their own inhouse chips just won’t be comparable. AMD is one of the only alternatives. The gap between Nvidia and the rest is on the contrary, is likely to grow. That’s their moat, their R&D.
BigTech and Cloud monopolies have spent billions of dollars on Nvidia’s systems, and the chipmaker has not kept up with the demand. Even while TSMC has expanded their capacity and are now doing the same in Japan.
Supply-demand in the global semiconductor supply-chain wasn’t ready for the Generative AI boom. In early 2024, Nvidia euphoria is a different kind of beast. The Biden Administration’s Chip Act doesn’t seem to be rolling out as planned or with the desired effect. It’s pushing China to innovate faster in the semiconductor industry than thought possible the by the Americans.
Once China catches up in semiconductors and hardware, on a level playing field, they will out-innovate the U.S. The U.S. has sought to ban AI chips from China basically destroying the global free market in the process leading to a lower global economic growth. For Nvidia too, China represents at least 20 to 25 percent of its market. In the future this could even be more.
Intel, which once reserved nearly all its chipmaking capacity for its own processors, is in the midst of a pricy gamble to transform itself into a credible contract-manufacturing rival to Taiwan-based TSMC, one that is unlikely to pay off as the U.S. talent and leadership cannot rival what’s happening in Asia, namely Taiwan and specifically TSMC.
Masayoshi Son and Sam Altman are too late to the party expecting they can solve the supply-demand dilemma in A.I. chips anytime soon. Nvidia and TSMC have monopolies so strong that it would take many years to break their dominance. If the Generative AI and productivity engine holds true, semiconductor stocks are likely to have bigger growth off of Nvidia’s meme-stock rally of recent months.
We need to fully appreciate this moment in the Generative AI enthusiasm of the cycle.
Nvidia is Nearly a $2 Trillion Dollar Company
Nvidia’s stock is up 63% YTD and it’s only February 25th, and this is after a significant rally last year. Nvidia’s earnings and guidance were strong enough to keep the AI spirit on the stock market alive. Nvidia reported another blowout quarter, with revenue more than tripling from a year ago — to more than $22 billion — and per-share earnings up 765%.
Nvidia's next-generation B100 products are based on the all-new Blackwell architecture that promises to significantly improve performance of AI compute compared to the existing Hopper architecture, according to the company's performance projections.
Even as the U.S. pushes its Generative AI first-mover advantage, it pushes China to innovate faster and become the future winner. Both Nvidia and AMD’s CEOs are Taiwanese-Americans and China is in the process of evaluating if a military invasion of Taiwan is feasible.
Nvidia it turns out isn’t just forcing markets to reevaluate AI hardware and data center advantages, but forcing the AI arms race to accelerate too. Given sheer demand for high-performance AI processors on the market, it is likely that Nvidia's existing customers have already pre-ordered at least some B100 products. Meta and Microsoft truly bought a lot of H100 chips, it’s hard to even imagine. Meta will reportedly own more than 340,000 of Nvidia's highly sought after H100 chips this year.
Since Nvidia supplies over 80% of AI chips of the highest quality, it can set the price for its B100 products. This gives it leverage in the short to medium term with revenues and margins soaring.
领英推荐
Nvidia became Synonymous as an AI chip leader in 2022
No longer was it about gaming GPUs, bitcoin or even data centers, Nvidia became the AI chip stock sometime in late 2022 to early 2023. That was the inflection point where we realized Nvidia had a monopoly on AI chips. We being the consensus of the market.
Loading video
Overall, global demand for semiconductors has been outstripping production. China can still access H100 and B100 chips off of a black market mostly via the Middle East. G42 is such an example of such an actor, even as OpenAI draws up a partnership with them.
Now the question remains in 2024 about how fast will Nvidia be able to ramp up production of B100 SXM modules and B100 PCIe cards as well as DGX servers on their base. How fast can Nvidia keep innovating against the entire world? But the semiconductor industry is extremely centralized and consolidated already. It took Nvidia decades to reach this point. Decades of failures, tribulations and real innovations.
AI chip giant Nvidia has rocketed in value so fast it passed Amazon and Google like they were nothing. Nvidia’s valuation rose the fastest and the most in history we’ve ever witnessed. If AI leaders are centralized in just a few names, the AI chip industry is even more centralized and this poses a danger for index funds, markets and hyperscale winner-takes-all Capitalism like America has created as the nucleus of its Global empire. The result of all that democracy and meritocratic capitalism in a formerly free-market has now resulted in a monopoly capitalism with surveillance capitalism characteristics. In such a world, advertising, the cloud and AI chips are everything.
Nvidia is a winner propped up by a system the US. is using to delay the fall of its vast empire of economic and technological dominance. Fed policy and banning rivals is how America competes on the world stage now, a kind of cold-war black tech strategy for global dominance in a new world of geopolitical tensions.
As such, Nvidia’s rise to power is a bit of a chimera. Generative AI hype isn’t translating into the productivity gains or useful applications like is being claimed by the markets and Silicon Valley, it’s another red herring. So what happens to markets when they find out? New technologies like LLMs and the race to AGI takes decades to mature, not months or years.
Nvidia however is a marvel of strategic leadership and they anticipated the 2023 rally as far back as 2017. That they now hyped expectations around its newest chip, the H200, which will ship in Q2 2024 with the promise of nearly double the performance of the AI business' current favorite, the H100 is an incredibly sticky narrative. Generative AI startups are popping up in the U.S. and China at an astonishing rate. National Governments will need to fund their own Sovereign AI policies thus enabling Nvidia to continue to expand.
Nvidia is investing in many of the world’s best AI startups as it supplies them with H100 and B100 chips. NVIDIA is also developing a variety of products, including the B40 GPU for enterprise and training, and the GB200, combining a B100 GPU with an Arm-based Grace CPU for training large language models. The H200 compute GPU, based on the Hopper architecture, is another addition, enhancing memory capacity and bandwidth for AI and HPC workloads. Despite supply chain optimization efforts, NVIDIA's CEO, Jensen Huang, has acknowledged the difficulty in immediately satisfying the demand for new products.
While Nvidia sold 2.5 million chips last year, Google spent $2 billion to $3 billion building about a million of its own A.I. chips, said Pierre Ferragu, an analyst at New Street Research. Amazon spent $200 million on 100,000 chips last year, he estimated. Microsoft said it had begun testing its first A.I. chip. BigTech should abandon in-house plans to build their own A.I. chips, and let Nvidia do its job. Reliance on Nvidia will be inevitable as they expand their R&D and partnership with TSMC, who are head and shoulders above everyone else.
Microsoft doing a custom chip with Intel won’t be able to compete. In fact customer chips for Nvidia will make it even more profitable in 2024 and 2025. According to Reuters, it’s a $30 Billion opportunity. From Microsoft’s Cloud emphasis to Meta’s Metaverse plans, the demand is not going to try up too quickly. Meanwhile Apple, Amazon and Alphabet need to attempt to stay relevant, though that’s somewhat in doubt.
As Nvidia has risen past a $2 Trillion market cap, it exposes some of the other hyper scalers of the so-called Magnificent Seven. With Google submitting its open-source SLMs via Gemma, Meta’s Llama-3 may not seem so incredible. Meanwhile Paris AI startup Mistral and its peers are making significant progress. OpenAI’s GPT-5 may not be the game changer GPT-3.5 or GPT-4 was to the evolving space.
Amazon and Apple in particular have a lot to prove and even how Google can protect its own moat. Their AI relevance is in question. Even in text-to-speech they have gutted their Alexa and Siri divisions respectively. Nor do they have the talent to compete even if they had as many AI chips as Microsoft or Meta. The majority of good talent at Google has already left for OpenAI and to found their own AI startups at this point. So while there is a consolidation in hardware, there is a diffusion in talent for another wave of Generative AI startups in the 2023 to 2025 period.
Of course this is all possible because of Nvidia and really because of TSMC. I hope whoever is POTUS in 2025 realizes this as Taiwan becomes more vulnerable to China’s military evolution and drone tech capabilities.
Belief and trust in AI is not really declining. After advertising and the cloud, we haven’t seen a new potential “cash-cow” like Generative AI for quite some time arrive on the internet. Nvidia and TSMC and to a lesser extent OpenAI are at the heart of a new multi-year revolution that won’t easily be a quick boom and bust cycle like many other failed experiments. Nvidia’s momentum in terms of supply-demand constraints and a lack of real competitors, is unprecedented.
While companies like Meta’s drive for efficiency and a dividend might make them a value trap now, there’s a lot of real innovation going on in AI that shapes the next generation of companies to enter the market. Unfortunately the GPUs are so costly, not many can keep up with the likes of OpenAI, Anthropic, Inflection and a few others. The hardware, AI agents and foundational model markets however can drive the AI narrative for years to come.
A Growing Total Addressable Market
What’s interesting is the TAM of some of these markets is now bigger than once thought. Estimates that the chip industry would grow to $1 trillion a year, once seen as aggressive, now look like they might be way too conservative, Gelsinger the leader at Intel. Today the semiconductor industry is about half of that, with 2023 sales estimated at around $523 billion globally.
What Nvidia calls “data center” revenue is now becoming something else entirely. And it might continue to exponentially drive new winners. Many of the stocks related to data centers and AI chips have seen huge increases that obviously do not make sense.
It is rumored that the Blackwell will be Nvidia's first architecture to adopt multi-chiplet designs. If this is the case, this may simplify production of Blackwell-based GPUs on the silicon level as it is easier to maximize yields of smaller chips. Nvidia’s leverage in 2024 is pretty incredible in the demand for these chips and I think it comes down to philosophy. The legend of Jensen Huang continues to grow:
Loading video
In 2023, Nvidia shipped 150,000 of its H100 chips to Meta, according to data from Omdia Research obtained by The Verge. Since Monopoly Capitalism is centralized in the hyperscalers, demand from the Cloud, advertising and E-commerce leaders will fuel the TAM of AI chips to unforeseen heights. This is basically what we’ll see in the 2024 to 2027 period.
The companies that stand the most to gain are the ones most key to the global semiconductor supply chain in 2024. To me these are companies like:
With ARM, Softbank have the most to gain from creating a direct Nvidia competitor. Masayoshi Son is rallying investors for a venture to challenge Nvidia’s dominance in AI chips, aiming to complement Arm Holdings with a $100 Billion AI chips venture. This one has the most credibility vs. the Sam Altman project in my opinion.
If Meta will have amassed 600,000 GPUs by the end of the year, other hyperscalers like Microsoft, Apple, Amazon and Google will also want to keep up. Finally Chinese BigTech will want to keep up, which means the black market for AI chips will become so lucrative it will created a boom in the Middle East for funneling them to China under the table. Geopolitics and AI thus collide in the 2020s in a unique context of the West vs. the East. Companies like TSMC are sort of neutral and agnostic if you think about it. They need customers in China but also have their HQ in Taiwan which is more politically aligned to the West, at least in theory.
The rise of Nvidia is one of the biggest stories in AI in early 2024 for a number of reasons. I will be even starting a new Newsletter called Semiconductor Things to cover these stories. I will announce it soon. It will be on this channel.
Markets are becoming so concentrated in just a few names in the magnificent seven it’s skewing everything. The S&P 500 is already up 7% so far in 2024. The NADAQ 100 is up nearly 8.5% YTD, these are like gains for an entire good year, in just 60 days of 2024. At all-time highs the stock market is riding an AI boom. But what would occur when inevitably the demand for data center and AI chips falls? I guess see you in 2026.
Markets are Becoming Too Centralized (rewind to 2014)
We could do a thought experiment about how fast things go in markets in just ten years with Fed stimulus of the pandemic era.
Magnificent 7 Market Caps, 10 Years Ago:
But is all of this even real? Index funds, boomers retiring and Fed stimulus skews the future.
Magnificent 7 Market Caps, Today:
How Monopoly Capitalism is Converging with AI chips
This is why they sometime say that Nvidia is eating the markets. Demand for Generative AI might not even be sustainable in the long-term as consolidation occurs more rapidly than in previous (GPT) general purpose technology cycles.
In A.I chips, with the partnership between TSMC and Nvidia, consolidation may have occurred before supply-demand issues were even fixed and fully addressed. Even as OpenAI got more money than its peers years before they could catch up, this creates new levers within America’s own monopoly tech strategy to dominate the world. Too much leverage of BigTech will mean more global opposition to America politically and geopolitically.
While Nvidia is a U.S. company, it’s ties to Asia cannot be denied. The intersection of AI and geopolitics has never been stronger and we stand at one of the most critical junctions of AI supremacy to date.