Future of Robots on Display at CES

Big show today from CES with fresh news beyond the floor, and markets easing after a strong tech run as traders wait on payrolls; feels like a little sell-the-news.

China is set to allow limited imports of Nvidia’s H200 for strictly commercial use, which knocked the stock early; the ban on military and state buyers still stands.

ByteDance and Baidu are eyeing as many as two hundred thousand H200s each, but Beijing’s approvals and US Commerce licenses must align; Jensen is optimistic and AMD has skin in this game too. China is a roughly fifty billion dollar opportunity for Nvidia even as local chipmakers gain ground and officials push buyers toward domestic alternatives.

Valuations keep stretching: Anthropic is seeking around ten billion dollars at a three hundred fifty billion dollar mark, with eyes on eventual public listings.

Investors frame the model race as four big players—OpenAI, xAI, Anthropic, and Google—with room for maybe three at scale; Anthropic’s edge shows in enterprise and code tools. Alphabet even leapfrogged Apple in market cap, a momentum shift few expected a year ago.

Part of Anthropic’s pitch is technical depth and a safety-first ethos from its OpenAI split.

Here at CES, humanoids stole the spotlight; Atlas drew two-hour lines, so we asked OpenMind’s Jan Lippart for a reality check.

A few years ago the bots looked like lab projects; now they look like real products. Software isn’t “solved” broadly, but logistics is mature while social interaction remains the frontier.

Plenty of builders tout vertical stacks; where does OpenMind fit across the 14 humanoids on display?

About half lean on our fast-moving software because many hardware-first teams don’t want to chase the frontier on both fronts.

Do we even want humanoids at home, or are softer forms better for daily life?

Humanoids click with human spaces—doors, switches, stairs—and my kids see ours as normal, while my mom was stunned by how far they’ve come. That compatibility is the point.

Can humanoids truly address elder care or is this more hype?

In memory care, patients bond so much that nurses wipe lipstick off robots at night, which shows how meaningful these interactions can be where human contact is scarce.

And your business model?

We want robots to be open like smartphones, with an app layer that lets developers build and extend instead of sealed black boxes.

Boston Dynamics’ Atlas is moving to production with Hyundai—here’s the vision.

By 2028 Hyundai will mass-produce between ten and thirty thousand humanoids a year, scaling to tens of thousands by 2030. Tesla is a serious competitor with a natural in-house customer in automotive.

On wearables, X-Real launched its entry AR glasses at four hundred forty-nine dollars and we sat down with CEO Chi Xu.

You’ve raised fresh capital—how much and from whom?

About one hundred million dollars from supply chain partners, with more details coming. I think glasses are the ultimate AI device because the display sits closest to your eyes and assistants can perceive what you see and hear.

Meta is pushing harder too; where do you stand in that race?

The category is big enough for multiple winners, and we just deepened a multi-year Android XR partnership with Google, naming us lead hardware partner to co-polish the end-to-end stack.

Will this look like Android phones with many OEMs and Google’s own hardware, and when do we see rollout?

Expect an ecosystem approach with multiple hardware makers alongside Google’s devices, and we’ll share go-to-market details when it’s ready.

How do you answer cost, comfort, and field-of-view questions versus premium headsets?

We’re targeting lighter, more affordable glasses that deliver most of the experience, plus real-time 2D-to-3D conversion on our own chip. Better yields, higher volumes, and a globalized supply chain drive the lower price.

Beyond CES, deal chatter around Paramount-Skydance versus Netflix, and Samsung’s memory-fueled profit surge, underline how central memory has become.

And with that memory squeeze and Jensen’s remarks still echoing, we turned to Palmer Luckey for a candid take on policy and competition.

There’s talk of a five million dollar CEO pay cap—what’s your comp and perspective?

I take one hundred thousand dollars in salary and own a big stake, so this hits legacy structures more than founder-led builders; I’m in the arena to fix problems even if critics say I’m conflicted. The rules apply to me too.

How do you view China’s lead in drones and the Taiwan risk?

China dominates drones through policy and supply chain muscle, while the US struggles on costs; a DJI ban should catalyze domestic and allied alternatives, as Japan has shown with a fully local build.

We then asked State’s Jacob Helberg how Paxelica fits into the AI and chip race.

We’re rallying advanced partners to win on silicon and compute by driving innovation, growing market share, and securing supply lines, with implementation targeted for 2026.

Does exporting H200s and cutting Gulf deals risk eroding our edge?

Selective exports keep developers building on American stacks while we shield the most sensitive tech, and we’re filling demand so others don’t seize the vacuum.

But Chinese rivals have surged—can we claw back share?

That’s why Paxelica runs protection and expansion in parallel and pushes diversification away from single-country chokepoints.

What’s your CES to-do list?

We’re pressure-testing plans with builders and mapping infrastructure across logistics, fabs, refineries, capital, and energy while we add members.

Given the week’s tensions, is this peace through power?

Deterrence needs credibility, and this diverse coalition—from Singapore to Israel to the UAE—is about pragmatic co-investment that raises resilience for all of us.

On venture, we spoke with Steve Jang—Nuro, Lucid, and Uber are teaming up on autonomy.

Uber is rekindling self-driving with partners, and Nuro’s platform plus Nvidia compute brings level four capability that should appear in the Uber app in the Bay Area later this year.

Nvidia is also pitching a full-stack kit; how does that shape the field?

They’re kingmaking physical AI by packaging chips, models, and data so more players build—and buy more compute—while staying years ahead; across autonomy, humanoids, and models, expect consolidation to a few winners as inference demand overtakes training amid memory constraints.

Tech faded into the close as hopes for blockbuster CES upside met reality, even with talk of huge future revenue and a pivot to physical AI; the robots were the stars.

Fourteen humanoids—and plenty of cuddly companions—lit up the floor, alongside real deals and fundraises to kick off the year; stay tuned for our Minimax interview on Hong Kong’s hot AI listings.

For the final time from Las Vegas, that’s Bloomberg Tech; recap on the pod, you know where to find it.