The Morning Download: Another AI Race

Good morning. It's on. OpenAI is laying the groundwork for a public listing in the fourth quarter of this year, people familiar with the matter tell the WSJ, while Anthropic has told financial partners it's open to going public by year-end.

Both companies are losing billions annually as they build and deploy frontier models and power their existing products. The Journal notes that going public could give whichever moves first a chance to tap a broader class of public-market investors. 

San Francisco realtors are sure to win no matter what. 

Enterprise customers might, too. Public filings would finally show what it actually costs to run these models, providing transparency that could reshape long-term vendor strategy and pricing expectations.

OpenAI may already be in a pre-IPO round. It is seeking to pull together more than $100 billion at a valuation of $830 billion. SoftBank is discussing investing about $30 billion, the Journal reported, and OpenAI has also discussed an Amazon.com investment of up to $50 billion. OpenAI is also seeking capital from Middle Eastern sovereign funds and major VCs.

But a year-end IPO timeline is ambitious. The WSJ notes:

The company recently has made changes to leadership ranks and is contending with fierce competition to its core consumer business from Google, prompting it to declare a weekslong code red effort to improve the quality of ChatGPT. OpenAI is also headed to trial in a case brought by co-founder Elon Musk, who is seeking up to $134 billion in damages.

And calling this a “race” may be a bit too much. For one, there are already too many: The AI talent race, the AI adoption race, the AI model-buidling race, the AI chips race, etc., etc. 

Model leads evaporate quickly, and enterprise adoption moves in fits and starts. Taken together, the AI landscape is less a sprint to the finish line than an endurance contest. 

WSJ Leadership Institute: How do you prioritize across all the different departments you oversee?

Subaiya: I've got great leaders across the board and all these different functions. They're pretty much self sufficient, but at the same time, there are things that bubble up.

With the whole SSD as well as a DDR RAM constraint within the industry, it's consuming a lot of time. Any change that needs to happen within the supply chain… takes a year or two because it has to go through multiple levels of approval. You cannot just jump start manufacturing in a new country.

WSJLI: You recently set up an internal-focused AI office for all of Cisco and appointed a chief AI officer. Give us some examples of what this department does.

Subaiya: Number one is getting the right infrastructure in place, which is backed by a workflow that ties into your existing workflow. One of the biggest things that I've seen a lot of companies make the mistake of is creating parallel workflows and not integrating AI into their existing workflows.

Next is applications. I typically look at it in four separate ways: Improve sales productivity, improve engineering productivity, improve customer partner productivity, and then there's the overall employee productivity.

WSJLI: How does your AI office drive ROI in those areas?

Subaiya: You need to… get rid of auxiliary functions. This is where you depend on contractors, where you depend on external vendors, you need to be able to reduce that, or even eliminate that.
You also need to look at the applications and say, which one of these can become automated workflows.

And the last one is pure engineering productivity gains. And this is not using AI development tools to do your [quality assurance] for you. This is actually using them to write your code for you.

Memory crunch coming for our Apple devices. Thanks, AI. Chief Executive Tim Cook Thursday hinted at higher costs possibly affecting Apple’s future financial results, the result of AI companies gobbling up capacity, the Journal reports. Chipmaker TSMC has so little spare capacity that Apple’s share of its revenue has dropped. Makers of Flash and DRAM chips, important components for devices like iPhones, are also in heavy demand by AI companies.

The Journal writes:

Apple gave little hint of its own plans to manage the memory crunch. But the company did project gross profit margins picking up slightly to 48.5% in the current quarter, suggesting confidence that it can keep sales and profit up in the near term. It needs to.

If it's any consolation, PC gamers are also feeling the hurt, the New York Times reported.

Anthropic's terms and conditions run into Pentagon ambitions. The company's models and coding tools are surging in popularity. At the same time, Chief Executive Dario Amodei has been very outspoken about AI's risks: “I don’t think there’s an awareness at all of what is coming here and the magnitude of it,” he told the WSJ last week.

That mix—a hot AI startup, a CEO both “excited and worried” about AI's impact —could now jeopardize a $200 million contract with the Defense Department, the WSJ reports. Anthropic’s terms and conditions dictate that Claude can’t be used for any actions related to domestic surveillance. It also objects to having its technology used in autonomous lethal operations.

The WSJ reports that some administration officials are frustrated with the company setting boundaries on how its technology could be used, even for activities the government considers lawful and standard practice.

All the recent job cuts. Believe it or not, much of the current wave traces back to the hiring binge of the pandemic years, according to the Journal. Tech and logistics, two sectors that staffed up most aggressively in 2020 and 2021, are now seeing the steepest pullbacks.

Consider:

• Amazon.com on Wednesday said it would cut 16,000 corporate employees after cutting 14,000 workers in the fall.
• UPS on Tuesday said it expected to slash 30,000 jobs this year, on top of 48,000 previous job cuts.

What about AI? That too. We may be at the front edge of an AI-driven labor correction, economists and CEOs, tell the Journal.

On Thursday Dow announced that it would cut 4,500 employees as part of a cost-saving program aimed at employing AI and automation to reduce expenditure.

More to come. Economists at Goldman Sachs estimate that in 2026 AI will be responsible for as many as 20,000 net job losses per month, according to the Journal. 

$85.3 billion. Revenue from iPhones in the December quarter, 23% higher than the prior year. Sales were particularly strong in China, where Apple’s results have been uneven in recent years, the WSJ reports. 

$463 billion. Loss in Oracle market value since it hit a record on Sept. 10.

6% to 7%. The percentage of current jobs that could be displaced by AI according to a recent Goldman Sachs report.

Robotaxi crash. U.S. regulators are launching an investigation into Waymo after one of the company’s robotaxis hit a child during drop-off hours near an elementary school in Santa Monica, Calif. The National Highway Traffic Safety Administration said Thursday that the child ran across the street from behind a double-parked SUV before being struck by the autonomous vehicle, sustaining minor injuries. The incident occurred on Jan. 23.

SpaceXX? Elon Musk’s SpaceX is aiming to do an IPO as early as this summer with the aim of raising more than $1 trillion, the Journal reported last week. Leading up to the IPO, the rocket maker might have a wing man. Bloomberg reports that SpaceX is considering a tie-up with Musk's AI company, xAI, a potentially synergistic move given Musk's current obsession with launching data centers into space. 

AI enters the “facial skin micro movements” phase. A recent Apple acquisition – nearly $2 billion, one of its largest–is Q.AI, an Israeli startup that makes technology that can analyze facial expressions. Q.AI patents showed its tech being used in headphones or glasses, using “facial skin micro movements” to silently communicate, the FT reported. The deal could give Apple customers the ability to “have private, non-verbal discussions with an AI assistant,” the FT said.