| In this edition, the government’s planned stake in Intel might not be enough to give the US an edge ͏ ͏ ͏ ͏ ͏ ͏ |
|  | Technology |  |
| |
|
 | Reed Albergotti |
|
The US government’s plan to convert more than $10 billion in grants for a 10% stake in Intel is only a partial, temporary fix to the country’s reliance on Taiwan for advanced semiconductor manufacturing. Without the equity stake, Intel may never have seen the much needed cash. But the bigger benefit is the signal that Washington wants to save the chipmaker. A lifeline for Intel could be viewed as the chip equivalent of the strategic oil reserve — a way to protect against geopolitical forces that could cripple the US economy. But it’s time to accept that Intel is probably not going to fully bounce back and challenge TSMC. What the US really needs is a new champion of chip manufacturing that doesn’t just challenge incumbents, but leapfrogs them. For all the Trump administration’s hyperbole about making America great again, there’s been surprisingly little talk of winning the semiconductor race with a knockout punch. Instead, it’s been mostly about life support. Even a fraction of the $4.5 trillion tax cuts estimated over this decade could jump-start the US’s efforts to become the world’s most advanced semiconductor manufacturer once again. The slash-and-burn approach to funding academic research counters that goal. Keeping Intel alive makes a lot of sense in the short term. But if the US doesn’t produce the next TSMC, the country will pay a big price in the long run. ➚ MOVE FAST: Playing games. Forget AI data centers, NVIDIA is making a big push in renting out gaming GPUs. The cloud gaming service GeForce NOW will get major upgrades, the company announced Monday, potentially changing video game economics. ➘ BREAK THINGS: Playing with fire. NVIDIA is also making China hawks angry with new AI chips destined for the rival nation, thanks to President Donald Trump’s decision to loosen export restrictions on advanced semiconductors. |
|
Laure Andrillon/ReutersThe last time investigative reporter Jeff Horwitz took on Meta (then called Facebook) for how it affects teens, there wasn’t much that could legally be done about it. Social media companies enjoy extremely broad protections from liability for any content created by their users. The latest Horwitz bombshell — revealing an internal Meta memo that appears to allow the company’s chatbots to have “sensual” conversations with underage users — is different. The prevailing opinion of legal experts is that AI chatbots do not enjoy the same broad protections because they are not user generated. They are the same thing as the company communicating directly with its users. But this hasn’t yet sunk in for Silicon Valley companies. That will happen when the lawsuits start getting filed. It’s expected that social media companies will fight for the same protections for chatbots under Section 230 of the Communications Decency Act, but the leaked memo is a reminder that this is new territory when it comes to AI. On Tuesday, a group of US senators sent Meta a letter about the memo, Semafor scooped. This time, the company could face a lot more than a public grilling from politicians. |
|
The votes are in. Trump’s renewed pledge to ban mail-in ballots in US elections lacks an alternative plan and authority — this power lies with Congress and state leaders — but it does bring into focus how technology could rise to the challenge. Shai Bargil, the CEO of Israeli online voting platform Sequent, believes the US is only a few years away from offering mobile device voting for overseas individuals and voters in major cities. It’s one of the few issues that may bring the left and right together. Democrats want expanded voting access to underserved communities, and Republicans want authentication methods. Sequent uses a cryptographic system, which encrypts responses and uses mathematical proofs to ensure votes are counted correctly. Users vote through a website, where they also verify their identity using biometric or multi-factor authentication methods. The code is open source, and the vote count is auditable. Courtesy of SequentIn May, 1.2 million overseas Filipinos cast their ballots in the Philippines’ national election using Sequent’s voting system. More than 200 municipalities in Ontario, Canada, will offer online voting through Sequent in 2026, Bargil said. He expects to begin piloting it for overseas US citizens next year. — Rachyl Jones |
|
 CIOs and IT leaders don’t just follow trends — they drive them. That’s why IT leaders at global organizations subscribe to CIO Upside. Get exclusive insights on the innovations and strategies shaping the future of tech leadership — subscribe for free today. |
|
Fabrice Coffrini/AFP via Getty ImagesPirates of the AI seas: We’ve written about Anthropic’s copyright case against book authors in the past. Anthropic convinced a federal judge it was right on perhaps the most important legal fear for makers of AI chatbots. It’s allowed to use copyrighted works to train AI models. But it may face a trial on a much more basic allegation: That it didn’t bother to pay for the many books it used for training. Since the ruling, a big question remains as to why Anthropic, one of the most heavily funded startups in history, didn’t just buy the books. We now have a hint. According to court records, Anthropic CEO Dario Amodei cited “legal/practice/business slog” as a reason. The decision could end up costing the company a lot of money, as this write-up in Transformer highlights. As a book author, it’s never bothered me that some words I typed could help AI models get smarter. But it’s always irked me that people on the internet can steal the book instead of buying it. From Anthropic’s point of view, slowing down even slightly to work out how to purchase perhaps millions of books would have been too costly for the company. Lawyers will now decide how costly that freeload was. Speaking for myself, I just hope most of the money goes to authors. |
|
 The amount of AI pilot programs employed by businesses that are generating little to no impact on revenue, according to new research by MIT’s decentralized AI program and reported by Fortune. While business leaders tend to blame regulation and limited model capabilities for their AI projects underperforming, MIT highlighted flawed integration by executives. |
|
Nvidia/Handout via ReutersLightning AI, the Nvidia-backed AI infrastructure startup, is launching a sui generis GPU rental marketplace that allows customers to compare processors from hyperscalers like Google alongside neoclouds like Lambda, the company exclusively told Semafor. Companies shopping for GPUs have traditionally had to choose between the two. Cloud providers like Microsoft, Google, and AWS offer graphics units for rent and machine learning services that help streamline training and deployment, but typically at a higher cost than the emerging neocloud sector, which includes firms like CoreWeave and Lambda. Cheaper neoclouds don’t have the same additional tooling as Big Tech firms, requiring customers to build their own infrastructure on top. Besides offering a side-by-side comparison — which customers could previously piece together with some legwork — Lightning provides its own machine learning platform for training and scaling models, mirroring what hyperscalers provide and closing the gap between competing providers. |
|
| |