All that, and a bag of chips

Welcome to Runtime! Today: why ridiculous sums of money are being discussed when it comes to chips and AI, the FTC puts AI vendors on notice, and the quote of the week.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

You know what would be cool

There is no question that the world would like to find multiple ways to ensure that a steady supply of chips will be available to fuel the generative AI boom. However, like an awful lot of conversations that have touched upon AI over the last year, people are getting ahead of themselves.

OpenAI CEO Sam Altman is supposedly in talks to raise trillions of dollars to fund "a wildly ambitious tech initiative" to build a network of chip-making plants around the world, according to a Wall Street Journal report from last week. On Friday, Bloomberg reported that Masayoshi Son, who is having a good month but has also lit billions of dollars on fire chasing earlier tech frenzies, is looking to raise $100 billion to compete head-on with Nvidia, building on Softbank's earlier investments in Arm.

Altman's venture would supposedly involve the United Arab Emirates and require as much as three times more the annual budget of the United States, which spends a lot of money.
It will also require the approval of the U.S. government, and might be just a trial balloon: "(Altman) is looking to the market for signals to determine whether to focus on a less extensive effort to build lower-level chips and software, or aim for a massive overhaul of chip manufacturing capacity," Bloomberg reported Friday.
Despite the report, building chip factories won't require actual trillions, and this headline in The Information made me laugh.
Meanwhile, according to Bloomberg, we learned that "Son has for years foretold the coming of AGI in his presentations, saying that a world filled with machines that are smarter than humans will be safer, healthier and happier."

Every tech breakthrough starts at the chip level, but pandemic-related supply-chain disorder and the incredible demand for Nvidia's AI chips exposed how reliant the world is on TSMC, which sits about 100 miles away from a country that doesn't think Taiwan should exist as an independent nation.

One of the goals of the Biden administration's Chips Act was to stimulate domestic production of semiconductors, but such investment was always going to take years to ramp up given how long it takes to build a modern fab.
Intel isn't expected to open its massive Ohio complex until 2026, even as it prepares to receive up to $10 billion in government subsidies.
And manufacturing is only one part of the process; Taiwan has long dominated the chip packaging industry, which is an essential step toward the end of the process and not one easily duplicated.
What Altman and Son are proposing would move the needle, but there's no guarantee that the money and chip-design expertise found readily throughout tech will translate into chip manufacturing, which is arguably one of the most difficult things we've figured out how to do as a species.

However, something needs to give if the generative AI boom is going to have the long-term effect on technology that its backers believe it will. While it's very early, it's not clear that enterprise investments in AI are paying off.

That's not to say that those investments have been wasted, but more that the current price of developing AI technology is out of line with its benefits.
Some companies will be happy to make those investments in order to learn how to best leverage AI and stay a step ahead of their competitors, but that doesn't make sense for everyone.
Chips are an enormous part of the expense required to develop and maintain world-class AI, as Altman obviously knows.
Still, it will take more than a network of chip-making facilities to beat Nvidia, which owns this market in large part because of its investments in software.

A MESSAGE FROM CANVA

Looking to integrate AI into your enterprise but not sure where to start? Join experts from Canva, Google Cloud and Ecosystm on February 26 at 4pm EST to discuss the hurdles of AI integration and what IT leaders can do to overcome them. Secure your spot for this webinar today.

Imposter syndrome

The FTC is considering whether or not to hold tech companies liable if their AI tools are used by impersonators to deceive consumers, it said in a blog post Thursday.

There are already rules preventing, or at least discouraging, people from impersonating others to pull a fast one, but the FTC wants to extend those rules to include AI tools. It also wondered "whether the revised rule should declare it unlawful for a firm, such as an AI platform that creates images, video, or text, to provide goods or services that they know or have reason to know is being used to harm consumers through impersonation."

If enacted, the rules could make it much harder to offer a lot of enterprise AI services, such as ones directed at customer-support call centers; what happens when somebody complains that they were tricked into buying that extended warranty by an AI bot that seemed really nice? The proposed rules are aimed at con artists with a talent for deepfakes, but depending on how they are implemented could have a wide impact.

Quote of the week

"Every company in the world is now a software company. Your competitiveness as a business is directly related to your ability for your software engineering team to ship changes." — Tom Wilkie, CTO at Grafana Labs, echoing Marc Andreessen's famous line when describing why observability tech has become so important.

The Runtime roundup

Dropbox stock fell by 23% Friday after it warned investors that 2024 revenue would be lighter than expected.

Broadcom and Google Cloud struck a deal to let current VMware Cloud Foundation customers run the software on either Google's cloud services or their own on-premises servers.