This was the week copyright law bent toward big AI.

Meta beat the Authors Guild. Getty dropped its US copyright claims. OpenAI picked a public fight with the New York Times. And a new standard for AI data permissions quietly launched with no teeth and even less attention.

Disappointing or necessary?

You decide.

I’m Ben Baldieri. Every week, I break down the moves shaping GPU compute, AI infrastructure, and the data centres that power it all.

Here’s what’s inside this week:

Let’s get into it.

The GPU Audio Companion Issue #49

Want the GPU breakdown without the reading? The Audio Companion does it for you—but only if you’re subscribed. If you can’t see it below, fix that here.

Multiple courtroom and industry signals suggest that copyright protections around training data are eroding fast.

First, a US federal judge dismissed most of the Authors Guild’s claims against Meta, ruling that training AI on copyrighted books doesn’t necessarily violate the law. Next, Sam Altman publicly challenged The New York Times, accusing it of “hacking” ChatGPT to produce its copyright-infringing outputs. Getty Images also quietly dropped key copyright claims against Stability AI in its lawsuit in the US.

These shift reflects growing legal momentum behind the idea that training is not the same as copying.

It’s not all bad news, however.

Creative Commons launched “CC Signals”, a voluntary standard to help creators label how their work can (or can’t) be used in AI training. It’s a polite nod to consent in a landscape increasingly dominated by opt-out-by-default. Whether it will be ignored again by the crawlers remains to be seen.

Why this matters:

  • While there’s a pretty compelling set of arguments for copyright reform, these rulings seem to be laying the groundwork for mass AI training on copyrighted data, whether creators like it or not.

  • Unless regulators act with the end in mind, copyright law may simply evolve through precedent, and mainly in the direction of the frontier labs.

  • If that happens, it’s starting to look more and more like a classic case of regulatory capture: Big tech writes the models and the rules.

Read the full stories from TechCrunch on Meta, OpenAI vs NYT, Creative Commons, and Getty.

Anthropic Picks Tokyo for First Asia Hub

Anthropic is going east.

The safety-first AI lab just announced Tokyo as its first Asian outpost. This follows months of interest from Japanese corporates and government officials, who are keen to deepen national access to frontier AI. The company’s Claude models are already being trialled in Japanese, and Claude 3.5 is expected to support broader multilingual applications across sectors.

Why it matters:

  • This marks the start of Anthropic’s serious international expansion.

  • Japan is a logical first step, given its advanced economy, strong AI research base, deep concern about tech sovereignty, and active courting of the major AI startups.

  • Claude’s success in Japan could put market pressure on hyperscalers and neoclouds to offer Anthropic’s models on a wider basis.

Read the full story on Nikkei Asia here.

Nebius Targets Healthcare in Vertical Play

Nebius is going niche.

The “Ultimate cloud for AI innovators” has just launched its first AI Discovery Awards as part of its focus on healthcare and the life sciences. The aim: fund early-stage projects using Nebius infrastructure for drug discovery, genomics, clinical diagnostics, and synthetic biology. Winners get credits, mentorship, and access to Nebius’ full stack of AI tools and compute.

Why it matters:

  • Domain-specific AI requires tuned models, privacy tooling, and sector-specific infra that general-purpose clouds don’t always offer.

  • Expect other neoclouds to follow with vertical bets in finance, legal, manufacturing, or any other area where domain-specific requirements will yield wins.

  • There’s no hyperscaler moat if you go deep enough into a niche.

Read the full story from Nebius here.

DeepMind’s AlphaGenome Tackles DNA’s Dark Matter

DeepMind has launched AlphaGenome to decode the “dark matter” of DNA.

This new tool can analyse up to a million DNA base pairs and make thousands of predictions on biological behaviour, from gene expression levels to the effects of single-letter mutations. In early tests, AlphaGenome successfully identified non-coding mutations that activate cancer-driving genes in leukaemia. DeepMind is pitching this as an “all-in-one” foundation model for genomic science, similar to how AlphaFold changed protein research.

Why it matters:

  • This is a serious bid to create a foundation model for biology, not just a better genomics tool.

  • If it works, AlphaGenome could drastically reduce the time and cost of identifying disease-related mutations.

  • The long-term play here could be massive: unlocking treatments, improving personalised medicine, and reshaping drug discovery.

Read the full story in Nature here.

Texas Instruments Reshores with $60B Investment

Texas Instruments is making a serious bet on American silicon.

The company just announced plans to invest $60 billion through 2030 in expanding its US semiconductor manufacturing. This includes additional fabs at its Texas and Utah sites, focused on analogue and embedded processors. The commitment doubles down on TI’s domestic footprint, with much of the capacity aimed at automotive, industrial, and critical infrastructure markets.

Why it matters:

  • Analogue and embedded chips are essential for power systems, sensors, and control infrastructure in data centres and AI-driven industries.

  • AI needs more than just GPUs, and TI’s investment reinforces the US's ambitions to reshore domestic manufacturing capacity for all types of silicon.

  • Industrial rebalancing is the trend du jour, so expect more silicon players to announce US (and maybe even EU) fab plans as geopolitical pressures and AI continue to reshape supply chains.

Read the full story on Tom’s Hardware here.

Amazon Commits £40B to UK Cloud & AI Expansion

Amazon just raised the stakes in the UK.

The company announced it will invest £40 billion (~$50 billion) into its UK cloud and AI infrastructure by 2040. This will fund the expansion of AWS data centres, including a focus on sovereign cloud services, AI infrastructure, and zero-carbon energy projects. The plan also includes building out AI and cloud training programmes for tens of thousands of UK workers.

Why it matters:

  • The UK is emerging as a major battleground for cloud and AI sovereignty.

  • With the region increasingly questioning US tech reliance and multiple neoclouds actively challenging the hyperscalers for dominance in AI, this deal is as much about perception as it is about capacity.

  • With this, along with its distinct German legal entity, Amazon is playing politics to position itself as the hyperscaler of choice for AI-era compliance.

Read Amazon’s full announcement here.

11GW AI Campus Planned for Texas

A former Texas governor just unveiled what might be the most ambitious AI infrastructure project on the planet.

Rick Perry’s new company, Fermi America, is teaming up with the Texas Tech University System to build an off-grid 11GW AI and energy campus near Amarillo, powered by a mix of nuclear, natural gas, solar, and battery. The 5,800-acre site is designed as a behind-the-meter HyperGrid campus with 18 million square feet of data centres and proximity to America’s primary nuclear weapons facility. Fermi has already applied to build four Westinghouse AP1000 reactors and expects the first 1 GW to come online by 2026.

Why it matters:

  • The UK's total installed power generation capacity was 74.8 GW in 2023, meaning this campus is equivalent to ~15% of a G7 country's total capacity.

  • If the project is real, it will be the world’s largest AI data centre campus.

  • Texas is doubling down on its claim to be the engine room of American AI, with energy and politics aligned.

Read the full story on DCD here.

The Rundown

AI is warping the things we once assumed were permanent.

Copyright law? Being bent in real time to fit the training needs of trillion-dollar labs. Infrastructure? A single AI campus in Texas is on track to generate more power than some countries. Mysterious aspects of our biology? Being unravelled at scale and at pace. Even regulation, once a constraint, now seems to shift conveniently in line with corporate momentum.

Everything’s getting bigger.

Everything’s getting faster.

And the institutions meant to slow it all down are, more often than not, just trying to keep up.

See you next week.

p.s. I’ve created a private LinkedIn group for subscribers. Complete the 1-minute survey below for an instant invite!

Get Instant Group Access

Reply

or to participate