Sam Altman’s $7 trillion chip moonshot, the FTC probes OpenAI and Anthropic deals, and the EU AI Act clears an important hurdle

Sign up
Worth Knowing

OpenAI’s Sam Altman Wants to Raise $7 Trillion to Make Chips: OpenAI CEO Sam Altman has been in talks with investors to raise as much as $7 trillion for a project to dramatically expand global chipmaking capacity, according to a report from the Wall Street Journal. Observers handwaved the reported target — more than the market capitalizations of Apple and Microsoft combined — as an act of “showmanship” by Altman, but the story isn’t worth writing off entirely. The generative AI boom has been powered by massive amounts of computing power. Absent a breakthrough in AI training practices, the relationship between computing power and AI sophistication — the more compute, the better the model — means OpenAI will only be able to go as far as its chips can take it. The chip supply is dominated by a handful of companies — namely Nvidia on the chip-design side and TSMC on the manufacturing side — that are already struggling to meet demand. That puts Altman’s company in a precarious position. His reported discussions with investors, including government representatives from the United Arab Emirates and investors from SoftBank, are unlikely to bring in anywhere close to trillions of dollars. Even if they did, some observers noted, Altman would struggle to find the workers and other inputs needed to actually spend the money effectively. But it’s still noteworthy that Altman is trying to find ways to satisfy his company’s massive computing power appetite that aren’t wholly reliant on Nvidia.

Expert Take: “Sam Altman’s fundraising plans are incredibly ambitious and likely will face a number of challenges. In particular, wrangling the motivations and interests of involved stakeholders, including governments, semiconductor fabrication companies, design companies, private investors, and customers will be a key obstacle. Additionally, the geopolitical implications would be significant as involvement from U.A.E.-based investors would likely draw scrutiny from U.S. lawmakers and government officials. That being said, it is too early to say to what extent Altman will be able to accomplish his goals, as there are still many outstanding questions.” — Hanna Dohmen, Research Analyst
Chinese Chipmaker Has Near-State-of-the-Art Chip in the Works: China’s largest chipmaker, SMIC, expects to deliver an advanced 5-nanometer smartphone chip later this year, according to a report in the Financial Times. When the U.S. government announced controls on semiconductor manufacturing equipment exports to China in 2022, it targeted SME capable of producing chips at the 14 nm process node and below — several generations behind today’s leading-edge chips. As we covered in September, a Huawei phone running on a SMIC-made 7 nm chip was cause for concern in Washington, because it appeared to blow past the 14 nm threshold. Like that chip, however, SMIC’s 5 nm chip won’t be the result of export-control circumvention or a significant Chinese technological breakthrough. According to the Financial Times, the chips will be made using SMIC’s existing stock of pre-control imported equipment, which use older technologies than the state-of-the-art tools employed by chipmakers like TSMC, Samsung, and Intel. SMIC’s ability to squeeze such advanced chips out of older machines is impressive, but it has apparently come at a cost. FT reports that SMIC’s yield rate — how many usable chips it can produce per silicon wafer — is a third of TSMC’s, forcing SMIC to charge as much as 50 percent more than TSMC does for its 7 nm chips. With the United States and its allies continuing to block access to the most advanced SME, Chinese companies will find it difficult to produce 5 nm chips or beyond at anything close to the efficiency of chipmakers like TSMC.
The EU AI Act Survives A Scare with Member-State Approval Vote: The EU AI Act cleared a crucial hurdle in its path toward becoming law, receiving unanimous approval from EU member states earlier this month. As we covered in January, the February 2 vote had been shaping up as a potential showdown over concerns — voiced most prominently by French President Emmanuel Macron — about the proposed regulation’s impact on the EU’s innovation ecosystem. According to Politico, those concerns were sufficiently assuaged by the European Commission, allowing the act to move forward. Once made law, the AI Act will be among the world’s most powerful AI regulations, setting important new standards for AI development and deployment in the EU and likely influencing how it is developed abroad. The official text has not been officially published, but a copy leaked to Euractiv’s Luca Bertuzzi has been published online. After securing member states’ support, the act got the go-ahead from key EU committees earlier this week. It will still need final approval from the European Parliament during a vote expected to take place in April. While EP members could derail the regulation at that stage, insiders reportedly expect it to pass without much trouble.
Government Updates

FTC Looks into Big Tech Deals with OpenAI and Anthropic: The FTC announced it had launched an inquiry into investments in two of the world’s hottest AI upstarts — OpenAI and Anthropic — made by three big tech icons — Microsoft, Amazon, and Google. According to the FTC, it is hoping to gather information about the strategic rationale behind the investments and about how the transactions will impact market share and competition in the AI sector. While none of the investing companies holds a majority interest in the two AI labs, their investments have been significant — Microsoft has backed OpenAI to the tune of $13 billion, while Amazon and Google have committed $4 billion and $2 billion to Anthropic, respectively. The inquiry doesn’t necessarily mean that antitrust action is coming, but observers say it’s part of putting the companies on notice that their actions are on the FTC’s radar. The FTC isn’t alone in probing the relationship between OpenAI and Microsoft. Across the Atlantic, the UK’s Competition and Markets Authority launched an inquiry in December, and the EU is probing whether the relationship might need a closer review under its merger rules.

DIU Looks to Enter New Phase with “Strategic Effect”: In a report published earlier this month, the head of the Pentagon’s Defense Innovation Unit laid out his vision for the organization’s next phase, which he dubbed “DIU 3.0.” DIU Director Doug Beck’s report, published in partnership with the Center for a New American Security, outlines a plan to take the lessons learned during the DIU’s first eight years and “apply them with the focus, scale, and speed necessary to deliver the strategic effect required.” Established in 2015 (originally as “DIUx”), DIU was stood up to help the military adopt cutting-edge commercial technologies. As Beck’s report explains, the organization’s first eight years were largely spent laying the foundation: establishing relationships between the Pentagon and the tech world (DIU’s headquarters are in Silicon Valley) and proving that commercial tech can meet the DOD’s needs. Now, Beck writes, DIU needs to begin focusing on delivering “at the scale required for strategic effect.” The DOD’s Replicator Initiative — its “moonshot” plan, announced last August, to field thousands of attritable autonomous systems across multiple domains within 18 to 24 months — will be an important test case for DIU’s ability to deliver at that type of scale. That timeline was deliberately ambitious, but DIU officials say the initiative is on track to meet the targets set last year.

DOD’s AI Office Kicks Off a Public “AI Bias Bounty” Exercise: On January 27, the Pentagon’s lead AI office, the Chief Digital and Artificial Intelligence Office (CDAO), launched its first “AI Bias Bounty” exercise to identify bias and unknown risks in large language models (LLMs). The DOD has used bounty programs before to help identify cyber risks, but this exercise marks the first time it has applied the concept to AI. The crowdsourced effort is open to the public, and participants can earn a share of a $24,000 pot for their work. Last August, the DOD launched an effort — dubbed Task Force Lima — to “assess, synchronize, and employ” generative AI tools, including LLMs, across the DOD. The Pentagon says the AI Bias Bounty exercises are meant to help develop “novel approaches” to auditing and red-teaming AI systems. The current exercise will run until February 27, and a second exercise will be announced soon.

Commerce Takes Steps to Set up U.S. AI Safety Institute: Last week, the Commerce Department took key steps toward setting up the new U.S. AI Safety Institute. The AISI, initially announced last November, is tasked with leading the U.S. government’s AI trust and safety efforts, operationalizing NIST’s AI Risk Management Framework (AI RMF), and developing technical guidance for regulators. On February 7, Commerce Secretary Gina Raimondo named Elizabeth Kelly as AISI’s inaugural director and Elham Tabassi as its first chief technology officer. Kelly will take over the role after serving two years on the White House National Economic Council, while Tabassi is a long-tenured veteran of NIST, which will house AISI, where she played a key role in developing its AI RMF. On February 8, Raimondo announced the creation of the U.S. AI Safety Institute Consortium, comprising more than 200 companies and organizations, including CSET. According to NIST, the AISIC will help achieve some of the priorities laid out in President Biden’s 2023 AI Executive Order, including setting guidelines for AI red-teaming, risk management, and watermarking.

In Translation
CSET’s translations of significant foreign language documents on AI

PRC Future Industries: Implementation Opinions of Seven Ministries Including the Ministry of Industry and Information Technology on Promoting the Innovative Development of Future Industries. This Chinese government document identifies “future industries” in six areas — manufacturing, IT, materials, energy, space, and health — as priorities for China’s industrial policy. The document names quantum computing and brain-inspired technologies, in particular, as areas of focus, and assigns a prominent role to Chinese central government-owned enterprises in developing these future technologies.

Shenzhen AI Industry Plan: Certain Measures of the Qianhai Shenzhen-Hong Kong Modern Service Industry Cooperation Zone Management Bureau to Support the High-Quality Development and High-Level Application of Artificial Intelligence. This regulation describes various financial and policy incentives that Qianhai District in southern China’s Shenzhen City is taking to promote the growth of its AI industry. It is one example of the types of subsidies and other supportive policies that Chinese local governments take to attract and incubate AI companies, often in competition with other localities.

If you have a foreign-language document related to security and emerging technologies that you’d like translated into English, CSET may be able to help! Click here for details.

Job Openings

We’re hiring! Please apply or share the role below with candidates in your network:
  • Research Fellow – Workforce: We are currently seeking candidates to support our Workforce line of research as a Research Fellow. In this role, you will be a key player in helping us accomplish three major research tasks: updating CSET’s assessment of the U.S. AI workforce, executing an analysis of the U.S. cybersecurity workforce, and assessing the U.S. educational pipeline for AI, cybersecurity, and STEM talent. As a report to our Senior Fellow who leads this research area, you will play a key role in shaping research priorities, overseeing research strategy and its execution, and leading a talented team of researchers. If you’re passionate about technology policy and possess strong leadership skills, apply now to make a meaningful impact on the future of America’s Workforce. Apply by Monday, March 4th
Please bookmark our careers page to stay up to date on all active job postings. You can also subscribe to receive job announcements by updating your subscription preferences in the footer of this email.

What’s New at CSET

PUBLICATIONS

EMERGING TECHNOLOGY OBSERVATORY

EVENT RECAPS

IN THE NEWS

What We’re Reading

Paper: AI-Powered Patching: The Future of Automated Vulnerability Fixes, Jan Keller and Jan Nowakowski, Google (January 2024)

Article: Automation Does Not Lead to Leaner Land Forces, Jack Watling, War on the Rocks (February 2024)

Upcoming Events


What else is going on? Suggest stories, documents to translate & upcoming events here.