The Intel Bailout

Is Intel Too Big To Fail?

It is becoming difficult to go a week without a new, dramatic chapter in the story of Intel. This week was no exception, as a peculiar thing happened. The US government, a vocal champion of free-market capitalism, and SoftBank, a Japanese venture capital behemoth known for its wild futuristic bets, arrived at the same investment thesis: it's time to buy Intel. SoftBank is putting in $2 billion. The Trump administration is discussing taking a 10% stake.

Which raises the question: Why are they propping up the one chip company everyone had written off for the AI era?

The basic situation is that Intel is a mess. It is losing billions of dollars, its factories are years behind Taiwan's TSMC, and its product line is ill-suited for the AI applications. The $8 billion in government subsidies it was promised under the CHIPS Act has so far failed to solve the company's core problem, which is that it has a technology deficit and not enough customers who want to buy what it makes.

One way to think about the White House's move is that this isn't a bailout in the traditional sense of a cash injection. The plan, after all, is to convert previous grants into equity. Intel isn't getting a new check from the government; it's getting something potentially more valuable: a powerful new shareholder. The logic here is less about immediate financial aid and more about a strategic vote of confidence.

With TSMC in Taiwan producing nearly all of the world's most advanced chips, Intel is now effectively the US "national champion" by default. Its failure is a national security vulnerability. By converting grants to equity, the government isn't just handing out money for specific projects; it is sending a powerful signal that Intel's survival is now a matter of state policy.

That signal is designed to de-risk the company for private investors—like SoftBank—and assure the market that Intel will not be allowed to fail. It is a backstop for a company whose financial foundation is visibly cracking. Just this month, credit rating agency Fitch downgraded Intel's credit rating to BBB, two notches above junk status, citing the very technology and customer problems that made this government intervention necessary in the first place.

As the Wall Street Journal reports, the White House's grant-for-equity swap is seen as a way to achieve the rescue while protecting the initial investment:

One option under consideration is converting some of the money that Intel was slated to receive from the 2022 Chips and Science Act into an equity stake. Commerce Secretary Howard Lutnick is searching for ways to improve the return on investment from the funds granted to companies like Intel under the law, one of the people said. Lutnick believes that converting those funds into an equity stake in Intel might be the best way for the government to bolster the company while protecting taxpayer interests, the person said.

This sort of direct state intervention is part of a pattern. In recent months, the administration has structured a deal where Nvidia and AMD pay the US Treasury a 15% cut of their China revenue, and it has secured a "golden share" in U.S. Steel. What's happening is a shift toward a more explicit, transactional form of state capitalism, where bespoke deals are struck with individual companies to achieve political aims.

SoftBank's role in all this is particularly interesting. Its $2 billion check serves as a convenient private-sector validation for the government's unprecedented move. It allows the White House to frame its intervention not as a desperate rescue of a failing company, but as a savvy public-private partnership alongside the "smart money."

But the essential question remains. All of this financial engineering—the government stake, the foreign capital, the political deal-making—is designed to solve Intel's confidence problem, and by extension, its capital problem. It does nothing to solve its customer and technology problems. And you can't fix a broken factory by having the Treasury Department join your board of directors.

The Thirsty Cloud

One way to think about the AI revolution is that it is a process of dematerialization. You take physical inputs—factories, supply chains, human labor—and you replace them with digital code running in the "cloud." The reality, it turns out, is a bit messier. The cloud is not ethereal; it is a collection of very large, very hot buildings that require a shocking amount of energy to run and even more water to cool.

And so you get the interesting situation unfolding in Johor, Malaysia. The state has, in a few short years, become one of the hottest data center markets in the world, the physical manifestation of the AI boom. Tech giants from Google to Microsoft flocked there after neighboring Singapore, worried about the strain on its power grid, put a moratorium on new data centers in 2019. Now, Johor is learning the same lesson Singapore did: the AI boom is a very thirsty business. Here is CNBC on the scale of Johor's dilemma:

While Johor currently has about 580 megawatts (MW) of data center capacity, its total planned capacity — including early-stage projects — is nearly 10 times that amount, according to figures provided by data center market intelligence firm DC Byte. That energy capacity would be enough to power up to 5.7 million households an hour... Kenanga Investment Bank Berhad, a Malaysian independent investment bank, has projected that the electricity use of the country’s data centers will equate to 20% of its total energy-generating capacity by 2035.

This is, to put it mildly, a lot. The result is that Johor is now slowing down approvals for new projects, struggling to keep up with the physical demands of its digital gold rush. This local story highlights a fundamental, global problem. The IMF reported that in 2023, data centers already used as much electricity as Germany, and by 2030, the electricity usage will be comparable to that of India, the world's third-largest electricity user.

The AI build-out, then, creates a fascinating policy paradox. On one hand, you have places like Malaysia and Singapore trying to create green, sustainable data centre frameworks to manage the immense resource costs. On the other hand, you have the US, where Trump's "America's AI Action Plan" calls for actively removing environmental regulations to accelerate the construction of data centers and the power plants they need.

It seems you can have an unrestrained AI boom, or you can have climate goals, but it is becoming increasingly difficult to see how you can have both. The race for AI dominance is quietly becoming a much more tangible race for electricity and water, and the winning strategy is not at all clear. The one certainty is that someone, somewhere, will have to build a lot more power plants.

The Scoreboard

• Cybersecurity: Palo Alto's forecasts signal AI boost for cybersecurity tools, shares rise (Reuters)
• AI: OpenAI’s Sam Altman sees AI bubble forming as industry spending surges (CNBC)
• Consumer: Samsung taking market share from Apple in U.S. as foldable phones gain momentum (CNBC)