Nvidia and the AI Bubble: Is It Real?

Over the week, Nvidia (NASDAQ:NVDA) stock dropped by 5.2% to $138.25 per share, leveling to the first half of October price range. It is telling that this price correction occurred after another earnings per share (EPS) beatdown. On November 20th, for the Q3 ending October 27nd, Nvidia surprised investors by 11.43%, having beaten the EPS estimate of $0.7 at $0.78.

Year-over-year, Nvidia’s EPS was 111% higher, while NVDA stock got boosted 189% for the same period . Yet, the $2 billion revenue estimate beatdown to $35.1 billion (up 94% YoY) seems to have elicited muted enthusiasm.

On one hand, it makes sense there would be cashouts for such a continuously and rapidly rising stock. This would then serve as the buy the dip opportunity. On the other hand, is AI data center demand as high ahead as the last two years of NVDA stock performance indicated?

Is there Such a Thing as an “AI Bubble”?

For there to be a bubble, the sector would have to be massively overvalued. Not just in terms of demand but also in terms of AI products returning investments. AI deployment across text-to-text, text-to-image, and text-to-video may as well become omnipresent, but there is a lingering question if such deployment will be adequately profitable.

After all, the popping of the dot-com bubble in early 2000 did not cause the internet to linger in stasis. On the contrary, but the Nvidia of the dot-com bubble, Cisco Systems (NASDAQ:CSCO) is still far away from its all-time high price of 80.06 in March 2000, vs its current price of $59.38 per share.

According to Crunchbase data, AI startups have cumulatively raised over $150 billion since 2021. Notably, this capital growth is escalating, having grown 80% more in Q1’24 than in Q1’23. Per Stocklytics, this led to $33 billion raised capital in just the first half of 2024. For comparison, US-based VC funding for crypto startups was just over $7 billion (trailing three months) at the peak of 2022.

That was before a series of crypto bankruptcies (BlockFi, Terra, Celsius, FTX…) flatlined the crypto market, alongside the regulatory suppression via Operation Choke Point 2.0.

If we go by the largest AI funding beneficiaries, OpenAI, the company is expected to incur a net loss of $1.3 billion in FY24, as of NYT reporting at the end of September. Likewise, Anthropic, with Claud AI challenger to ChatGPT, is on track to lose $5 billion this year.

It turns out, scaling of computing costs, including the electricity, GPU acquisition and server maintenance, is a difficult problem to tackle. At a glance, this may point to an AI bubble. However, there are some key extenuating circumstances.

Concerted Effort to Make AI Happen

The crypto market received heavy suppression from all angles. This was predictable and explained by congressman Brad Sherman. In contrast, the AI sector has received institutional blessing to such an extent as to facilitate the recommission of nuclear reactors for Microsoft’s data center power needs.

The overarching goal is to utilize AI to automate the moderation of global content. The World Economic Forum (WEF), as the hub of public-private partnerships (PPPs), has been pushing this agenda through the Global Coalition for Digital Safety.

Likewise, the numerous “AI safety” meetings between politicians and businessmen (public-private fusion) are laser-focused on AI-powered algorithmic control.

“How you understand, master and harness this technology revolution will define the place of this country and the shape of the world,”

Tony Blair, former UK PM and the head of one the largest NGO complexes in the world, the Institute for Global Change (TBI)

Such an algorithmic control has already been evident in Microsoft (NASDAQ:MSFT) Bing’s aggressively censored image creator. This is in line with Microsoft CEO’s effort to limit “unintended consequences” of AI. In other words, just as the internet evolved from a decentralized space to account-based centralized platforms, so it will happen with AI.

But Nvidia is likely to be the main beneficiary of this process. So far, Nvidia’s full-stack approach in AI model training has made the company corner ~80% of the AI chip market. The bulk of this demand comes from Big Tech hyperscalers, as the established pillars of the centralized internet.

Of course, the “Big” part of the Big Tech infers inherent participation in WEF’s public-private partnerships (PPPs) push.

“Only coordinated governance can ensure that AI delivers benefits inclusively and ethically, especially during heightened geopolitical instability.”

Conversely, given the established PPPs in the finance sector itself, it is exceedingly unlikely there will be a capital withdrawal from the AI sector. Moreover, Nvidia is yet to benefit from the computing power needed for inference.

For particular queries, real-time AI inference is equivalent to human reasoning and data interpretation, going off unseen data. Upcoming Nvidia’s Blackwell (B200) architecture performance is at the top in the inference department. This is on top of the doubling of the LLM training performance.