Broadcom Follows Oracle in Disappointing AI-Focused Investors

This is Bloomberg Tech. Coming up, Broadcom stumbles after an AI-fueled run, China lines up a massive chip push, and we dig into the White House’s new AI order that targets state-level rules.

Markets are under pressure with tech leading the drop, while earlier strength in China-linked names faded as the session wore on.

Breaking: Oracle is pushing some OpenAI-focused data centers back to 2028, and the stock sank on the headline; the bigger story today is Broadcom’s slide after holding back a full-year AI revenue target despite touting a sizable AI backlog.

Broadcom actually beat and raised, called out AI revenue roughly doubling year over year, and guided next quarter’s AI sales above consensus, but investors are uneasy about margins tied to a large TPU system deal that assigns a significant slice of the AI backlog to a lower-margin configuration.

I’m still constructive and lifted my price target, because Broadcom is assembling a full-stack offering across networking and compute, similar to Nvidia, and hyperscalers will use both GPUs and custom silicon rather than picking a single path.

This is not an AI bubble; supply remains tight in components like optical gear, yet Broadcom signaled capacity to meet demand into fiscal 2026 and 2027.

China is considering the largest state-backed chip incentives yet, potentially tens of billions of dollars to shore up its domestic ecosystem.

We’re hearing a range of approximately 28 to 70 billion dollars that could flow to AI chip champs such as Huawei and Cambricon and to foundry player SMIC, though export curbs still limit access to advanced tools and it’s unclear how China will treat Nvidia’s H200.

Back to Oracle’s delay and why it matters for revenues tied to OpenAI workloads.

These are gigawatt-scale sites and Oracle is attempting several at once, so some completion dates slid by about a year; they’re turning capacity on in phases, like Abilene, but labor and materials are tight—especially skilled electricians in remote areas—so revenue recognition happens as each chunk goes live.

Does this drip of delays mean AI payoffs arrive later than hoped?

I see end-of-year nerves, not broken fundamentals; long-term AI demand looks intact and weakness in quality tech is a chance to add.

Big platforms can self-fund through heavy cash flow, while higher debt names like Oracle bear more risk; short-term bottlenecks are solvable, and next year should reward innovators over laggards.

Uber’s CEO is leaning into autonomy and Asia for growth.

APAC is a major engine for rides and taxis on our platform, and regulators across North Asia are engaging on pilot programs; we’re live with autonomous services in four markets now and aim for more than ten next year, with hopes to include Japan, Hong Kong, and Australia.

Uber and DoorDash are also suing New York City over a rule that would surface a default tip at checkout, arguing it adds to sticker shock.

Rivian is pivoting away from Nvidia for future vehicles by building its own chip and changing its sensor mix.

The stock surged as Rivian unveiled two big bets for the next-gen R2: an in-house AI processor made with TSMC and embedded lidar, moving beyond the camera-first approach; earlier models should gain hands-off capabilities via software, though scaling production remains the company’s biggest challenge.

Our new compute targets roughly 1,600 sparse TOPS and processes about 5 billion pixels per second with integrated lidar, cutting costs versus third-party chips and laying the groundwork for personal level four driving, like sending the car to pick up your kids.

Tech stocks are sliding into the close with AI jitters pulling even China names off their highs.

Now to the UK’s push to supercharge scale-ups.

We’re inviting US VCs to bring expertise while we unlock domestic pension capital, because the UK excels at incubation but needs deeper scale-up funding and networks to keep winners onshore.

Strengths include fintech and applied AI—think companies like Synthesia and ElevenLabs—and life sciences, and we’re raising a co-investment fund from UK pensions with a first close planned early next year.

Let’s turn to the White House executive order on AI and why it centers on state rules.

David, what problem were you solving, and why focus on stopping a patchwork rather than laying out one national framework now?

States are moving in different directions with hundreds of AI measures, which creates compliance landmines for startups; the EO sets federal principles and tasks the administration with working with Congress on a single national framework while equipping DOJ to challenge the most burdensome state laws.

Models are built, trained, and used across multiple states and the internet, so conflicting rules do not make sense; we want child safety, respect for copyright, and local say on infrastructure, but not fifty different algorithms of law.

How do you answer fears that AI will erase jobs and that Washington is catering to big tech?

So far we’ve seen job growth, not net losses, with sectors like construction booming and wages rising, and estimates suggest AI is lifting GDP; displacement may come in time, but the present picture is expansionary.

Would DOJ sue states like California or New York under this order?

A litigation task force can target overreaching laws—Colorado’s algorithmic discrimination statute raises First Amendment questions—but no specific state cases are decided yet.

And on Nvidia’s H200 in China—what happens next?

China appears to be shunning those chips to back Huawei and pursue semiconductor self-reliance, while the US permits only non-leading parts and keeps the latest architectures off limits.

That’s it for us as a bruising week for tech wraps with the Nasdaq down more than two percent and AI angst in focus.

Oracle’s data-center delay helped set the tone, and that’s a wrap on my final show of 2025—what a year; catch the podcast on the terminal and online, and Caroline, see you in 2026.