TikTok JV, AI Rulebook Wars, Memory Moats

Welcome to AI News in 10, your top AI and tech news podcast in about 10 minutes. AI tech is amazing and is changing the world fast, for example this entire podcast is curated and generated by AI using my and my kids cloned voices...

Here's what's shaking in AI and tech on Friday, December 19, 2025... TikTok finally inks a U.S. joint venture that could end years of ban threats. The Department of Energy rolls out a sweeping AI collaboration under its Genesis Mission. Capitol Hill heats up over who gets to write the AI rulebook. Australia backs away from a training-data carve-out. And Micron's blockbuster earnings remind everyone the memory squeeze isn't going away anytime soon.

Let's dive in.

[BEGINNING_SPONSORS]

Story one: TikTok, after half a decade of political whiplash, is taking a big step to secure its future in the U.S. TikTok and its parent, ByteDance, signed binding agreements to create a new U.S.-based joint venture with American and global investors — Oracle, Silver Lake, and Abu Dhabi-based MGX — tasked with overseeing U.S. data protection, algorithm security, and content moderation.

The deal targets a January 22, 2026 close. Reports vary on exact ownership stakes, but the throughline is consistent — American investors will hold significant control, ByteDance will keep a minority position near or below 20 percent, and Oracle will serve as the trusted security partner guarding U.S. user data. TikTok says more than 170 million Americans use the app, which explains the urgency to resolve national security concerns without a ban. Oracle shares jumped on the news in after-hours trading. According to reporting from Reuters, the Associated Press, and TechCrunch, the new entity — TikTok USDS Joint Venture LLC — will have a majority-American board.

A couple of open questions remain. How much practical influence will ByteDance retain through technical know-how... and who ultimately steers the core algorithm's evolution? Reporting indicates the algorithm will be retrained on U.S. data and audited under the new security framework — moves meant to address concerns about foreign influence and data access. Expect continued scrutiny from lawmakers even as this deal tries to thread the needle between national security and free expression.

Story two: A sweeping public-private science effort from the U.S. Department of Energy. The DOE announced new AI collaboration agreements under its Genesis Mission with two dozen organizations — Microsoft, Google, Nvidia, AWS, Oracle, Intel, AMD, HPE, OpenAI, Anthropic, xAI, Palantir, and more.

The goal is straightforward: accelerate scientific discovery across fusion and nuclear energy, quantum, robotics, and supply chains — while reducing dependence on foreign tech. Think of this as plugging frontier models into America's supercomputers and national labs, with cloud and accelerator muscle to match. Nvidia brings accelerated computing and models; Microsoft and Google bring cloud and tooling; OpenAI's program would route frontier models into national labs; Anthropic contributes Claude and an embedded team. The DOE is positioning Genesis as AI to scale science — not just chatbots, but materials, climate, and national security workloads. Reuters first detailed the scope and partners.

What to watch next — how fast these models adapt to domain-specific science tasks... whether the partnerships spur new open data standards... and how the effort balances security with scientific openness. Genesis builds on prior supercomputing collaborations, but the size, the breadth of partners, and the explicit focus on dual-use concerns put it in a different league.

Story three: A political showdown over who gets to write the AI rulebook. Senate Democrats are moving to counter President Trump's December 11 executive order that aims to preempt state AI regulations by establishing a single national policy. Senator Ed Markey unveiled the States' Right to Regulate AI Act, which would block federal funding to implement the order and defend state authority to set their own guardrails. Senator Brian Schatz is coordinating a companion effort in the House with Representative Don Beyer. Axios notes these bills face long odds, but they mark a serious pushback against federal preemption. At stake is whether state laws — like California's 2025 Transparency in Frontier AI Act and other algorithmic rules — get sidelined by Washington.

For context, the White House order argues that a patchwork of 50 different regimes would burden innovation and interstate commerce, and that some state rules risk imposing ideological biases or extraterritorial controls. The administration wants a minimally burdensome national standard. The response from Democrats suggests Congress will be the arena for this fight in 2026 — with the courts likely weighing in on the scope of federal preemption.

[MIDPOINT_SPONSORS]

Story four: Australia has pulled back from a contentious idea to let AI developers train on copyrighted material under a "fair dealing" exemption. After strong backlash from creators — particularly in music — the country's Productivity Commission withdrew the proposal. The government had already rejected the carve-out in October; the final report now recommends waiting three years before reconsidering whether AI requires new copyright exceptions, essentially saying the current regime is sufficient for now.

The argument against the exemption is straightforward: a broad carve-out could legitimize mass ingestion of creative works without consent or compensation. The global signal here is notable — while some jurisdictions flirt with training exceptions, Australia is slamming the brakes.

Practically, this means AI companies operating in Australia will need to keep negotiating licenses, rely on non-copyrighted data, or use synthetic and public-domain corpora — complicating model training at national scale. For creators, it's a victory in a year that included the U.S. TAKE IT DOWN Act and several state-level rules against deepfake abuse. For the rest of the world, Australia's move becomes a data point in the ongoing experiment of how — or whether — to fit generative training into existing copyright frameworks.

Story five: Micron just poured high-octane fuel on the AI infrastructure supercycle narrative. The memory maker delivered a monster quarter — about 13.6 billion dollars in revenue, well ahead of expectations — and guided the current quarter to a midpoint near 18.7 billion. The company says demand for high-bandwidth memory, the ultra-fast DRAM that feeds GPUs, remains well above supply and could stay that way for years. That sent Micron shares up more than 10 percent and lifted other AI-exposed chip names. Barron's called it some of the strongest upside seen in chips outside Nvidia's 2023 bonanza.

Why this matters beyond the ticker tape: high-bandwidth memory is becoming a second choke point after GPUs. Micron's commentary suggests HBM capacity for 2026 is essentially spoken for, and that gross margins are expanding as pricing power shifts to memory suppliers. That has ripple effects — cloud providers may face higher total system costs, smaller AI startups could get squeezed on training budgets, and PC and phone makers might see tighter DDR5 supply as fabs prioritize HBM lines. Multiple reports now frame the AI build-out as a multi-year — not quarter-to-quarter — phenomenon.

Rapid-fire look-ahead: with TikTok's joint venture on track for a late-January close, expect regulatory filings to reveal more about board composition, audit rights over the algorithm, and incident response protocols. The DOE's Genesis Mission should start surfacing pilot projects — keep an eye on materials discovery and grid optimization demos. On the policy front, watch whether the Senate bills force hearings on the federal-versus-state AI powers question, and whether other countries follow Australia's lead on training-data exemptions — or go the other way and formalize fair-use style carve-outs. And on the hardware side, Micron's signal implies sustained capex from hyperscalers and a need for more advanced packaging — another bottleneck to track in 2026.

That's a wrap: TikTok's U.S. joint venture moves ahead, the DOE bets big on AI for science, Washington gears up for an AI federalism fight, Australia taps the brakes on training exemptions, and Micron reminds everyone that memory is the new moat in AI. We'll keep watching how these threads — policy, platforms, and parts — intertwine as 2025 closes out... and we head into a pivotal year for AI.

Thanks for listening and a quick disclaimer, this podcast was generated and curated by AI using my and my kids' cloned voices, if you want to know how I do it or want to do something similar, reach out to me at emad at ai news in 10 dot com that's ai news in one zero dot com. See you all tomorrow.