The next wave of AI hype will be geopolitical. You are paying.

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

A popular view of creative AI is that it is unjustifiably expensive, chronically wasteful, rarely useful, and is being foisted on the general public for theoretical reasons, even though it would make those services worse. They depend on them. Governments are fully convinced of this.

So says Barclays:

The initial wave of AI is well under way, fueled by billions of hyperscale dollars. The concern with Nvidia, and ultimately with the second wave of the AI ​​ecosystem, is where the next pocket of dollars comes from once hyperscale capex can no longer shift to AI or grow meaningfully year over year. can

In recent months, we've seen a growing push by nations around the world to rapidly educate themselves and stay at the forefront of AI's powerful capabilities. In practical terms, this equates to public announcements by several countries (Saudi Arabia, Singapore, Germany, UK, India) of plans to spend hundreds of millions and even billions of dollars towards supporting the AI ​​hardware ecosystem. will go

The private sector AI cash burn is already of sovereign size. Amazon, Meta, Google and Microsoft will jointly invest. About 200 billion dollars This year, according to Bernstein Research.

These sunk costs require some sort of reversal by the time depreciation charges reach their income statements. Continued acceleration in capex growth depends on companies finding something that the public wants to buy. Their need for income may soon be urgent, and the outlook so far has not been encouraging.

But because political leaders care more about one-upmanship than ROIC, taxpayer-subsidized AI could grow rapidly even if the corporate bubble bursts.

Barclays estimated in a recent note that if the former nations matched the US's $4 billion in AI spending, in terms of GDP, it would add another $3.5 billion:

For Nvidia, this adds up to barely a month's worth of revenue. More important is the alternating cycle.

Barclays estimates that hardware purchases will be obsolete within two years. As AI hardware becomes larger and more expensive, aggregate annual autonomous spending could quickly exceed $25bn:

Overall, we see AI as the most powerful enabler of technological progress, as well as a major security threat as adversary nations increase capabilities, ultimately justifying our projected costs. does and gives us confidence that the numbers must be materially higher.

The US has taken an early lead because its government has been relatively enthusiastic about AI. A federal use case inventory was published in September that identified more than 700 potential applications, and earlier this month a Senate roadmap for artificial intelligence policy proposed a $32bn R&D budget.

Although such figures may prove anecdotal, there is little scrutiny of large expenditures. The Senate's roadmap does not include defense, which accounts for nearly all current US federal AI spending.

A study of government tenders published by the Brookings Institution in March found that the US Department of Defense is aggressively ramping up AI investments in 2022. In terms of maximum potential contract value, last year was just over $4bn of the $4.56bn in AI procurement spending. Brookings, the defense agency, calculates

Senate Majority Leader Chuck Schumer has said that the US AI defense budget needs to be increased nearly eightfold. The real purpose of all this investment will be confidential information.

Barclays says countries following the US's lead will want something in-house that is at least equivalent to OpenAI's GPT-4. Last year's best technology represents “at least a starting point for nations seeking to be at the forefront of AI for both economic and security purposes.”

Costs for such a rig would cost $600mn at current processor blade prices, again to cover interconnects, storage, power costs, etc. Such a setup will do nothing to intimidate the enemy. This requires staying on the bleeding edge of AI, which is a Very more expensive.

GPT-4 is said to have used 25,000 accelerator cards to train, while GPT-3 — released less than three years ago — required only 1,000. The grid below estimates all current construction costs in increments of ten thousand accelerators, or XPUs.

Estimated AI Accelerator Spend by Cluster Size © Barclays

If hardware cost inflation continues at the current pace, the cost of an optimal AI computing cluster could easily exceed $5bn, says Barclays:

Estimated Next-Gen AI Accelerator Spending by Cluster Size © Barclays

Barclays says the information arms race will grow so fast, only 15 countries can afford to participate. And there is no turning back for those who are able to pay, as “AI capabilities have become one of the most important, if not the most important, national initiatives globally”:

In our view, the global development of AI applications will undoubtedly become a national security issue no different than how the government views domestic chip production, and through the lens of the ~$39bn CHIPS Act passed several years ago. Under, we see a lot of scope. In government budgets for increased spending on new clusters and more advanced hardware once readily available.

Additionally, we believe that new AI/computing investment projects proposed by Saudi Arabia ($40 billion AI investment fund according to the New York Times), Singapore, Germany and even India will lead to stronger AI writing. Instead, it can pressure policymakers to act quickly. Investment plans in policy

So buy Nvidia, Barclays tells clients. With the associated pile of sanctions and antitrust risks, the stock may seem expensive, but public servants would know no better than to buy servers off the shelf:

We see NVDA as the biggest beneficiary of Sovereign AI because of its already dominant merchant AI accelerator install base, performance leadership, and developer community preference. We also see the autonomous AI market as a strong potential adopter of the company's full rack solutions. [ . . . ] Overall, we see the expected spending from Sovereign AI as incremental to the entire AI ecosystem, given government agencies' lack of engineering know-how and the resources required to assemble custom solutions around merchant hardware. , and thus believe it will also permeate the broader AI ecosystem.

And sure. why not. Once the trailing PE goes above 300x, anything goes.

Being an ESG darling was a big part of last year's Nvidia buy case. A year ago it was adjacent to the shitcoin bubble, and before that it was about most. Cyberpunk 2077 Frame refresh rate. It is now a buy as it is the de facto weapons supplier for World War GPT.

A common thread that connects Nvidia's customers and shareholders is that they don't know what they're buying, or why they need it, but believe they must have it. An international arms race for billion dollar boondoggles and tiger repellent stones would fit that description.

Further reading:
– 'Sell Nvidia' (FTAV)

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Leave a Comment