This Week in Tech 116

This week we explore:

NVIDIA's annual GTC conference kicked off Monday in San Jose, and the vibe was less "chip company" and more "civilization infrastructure provider." CEO Jensen Huang announced that NVIDIA is looking at $1 trillion in demand for its Blackwell and Vera Rubin hardware through 2027. To put that in context: NVIDIA reported $78 billion in revenue this quarter, up 77% year over year. It has now posted 11 straight quarters of revenue growth above 55%. The most valuable public company on earth is still accelerating.

But the headline number isn't the real story. The real story is the framing shift. At GTC this year, GPUs were almost an afterthought. Huang spent the keynote talking about "AI factories," "physical AI," and what he calls the full stack: compute, open models, simulation frameworks, and robotics platforms working as a unified system. NVIDIA unveiled Cosmos 3, a world foundation model that lets robots learn in simulated environments before they ever touch the real world. Dozens of humanoid robotics companies, including AGIBOT, Figure, and NEURA Robotics, are now building on NVIDIA's stack. The pitch is that every industrial company is about to become a robotics company, and NVIDIA wants to be the operating system underneath all of it.

The autonomous vehicle angle got concrete too. Huang announced that Uber will launch an autonomous fleet powered by NVIDIA Drive AV across 28 cities on four continents by 2028, starting with Los Angeles and San Francisco. Nissan, BYD, Geely, Isuzu, and Hyundai are all building Level 4 vehicles on NVIDIA's Drive Hyperion program. The chip is the least interesting thing NVIDIA sells now.

Read more - Fortune​​​​​​​​​​​​​​​​

OpenAI dropped two new models today: GPT-5.4 mini and GPT-5.4 nano. On the surface, these are just smaller, cheaper versions of the flagship. Under the surface, they're OpenAI explicitly laying out the architecture for how agentic AI systems will work at scale.

Here's the model: let GPT-5.4 handle the planning, judgment, and complex reasoning. Then spin up mini or nano as subagents handling narrower parallel tasks. Mini runs more than twice as fast as the previous GPT-5 mini, nearly matches the flagship on coding benchmarks, and costs $0.75 per million input tokens. Nano goes even cheaper at $0.20 per million input tokens. In their own Codex platform, mini only uses 30% of the GPT-5.4 quota.

The benchmark that matters: mini hit 72.1% on computer use tasks versus the flagship's 75%. For navigating interfaces and interpreting screens, mini is almost as good at a fraction of the cost. For teams building high-volume AI workflows, the math just changed. The question is no longer whether to use smaller models. It's figuring out which tasks actually need the big one.

Read more - Blockchain.news | OpenAI

You shipped it. It works. Tickets are resolving. So why are customers leaving?

Gladly's 2026 Customer Expectations Report uncovered a gap that most CIOs don't see until it's too late: 88% of customers get their issues resolved through AI — but only 22% prefer that company afterward. Resolution without loyalty is just churn on a delay.

The difference isn't the model. It's the architecture. How AI is integrated into the customer journey, what it hands off and when, and whether the system is designed to build relationships or just close tickets.

Download the report to see what consumers actually expect from AI-powered service — and what the data says about the platforms getting it right.

If you're responsible for the infrastructure, you're responsible for the outcome.

Get the data

This is one of those reports that sounds crazy until you read the reasoning, and then you can't stop thinking about it. Bank of America's research team published a forecast projecting that the global humanoid robot population will hit 3 billion units by 2060, surpassing the number of cars on the road. By that point, 62% of those robots, roughly 2 billion units, will be deployed inside people's homes.

The driver isn't novelty. It's demographics. Aging workforces in Japan, Germany, South Korea, and increasingly the U.S. have strained manufacturing and services for years. Wage growth in logistics, warehousing, and elder care has outpaced broader inflation. BofA's analysts make a simple point: you don't need a perfect robot. You need one that shows up, doesn't quit, and costs less than the workers you can't find. Annual shipments are projected to climb from 90,000 units in 2026 to 1.2 million by 2030, an 86% compound annual growth rate, steeper than the early EV market.

Chinese-manufactured humanoids carried a bill-of-materials cost of $35,000 in 2025. BofA projects that falls below $17,000 by 2030. Western-built units currently cost $90,000 to $100,000 to produce, meaning the compression ahead is enormous.

This is the decade when the humanoid robot stops being a demo and starts being a product category. The question is who builds the platform everyone else builds on.

Read more - Fortune

There's a graveyard of startups that have tried to build a household humanoid and failed, usually because teaching a robot to handle a wine glass differently than a dish towel requires absurd amounts of training data. Sunday, founded by Tony Zhao and Cheng Chi, thinks they've cracked enough of the data problem to try.

They raised $165 million in a Series B led by Coatue, with Tiger Global, Benchmark, and Bain Capital all participating, at a $1.15 billion valuation. Their robot, Memo, targets laundry and table clearing as the opening wedge: repetitive, high-frequency household tasks that people hate enough to pay someone else to do.

They already have 1,000 people on the waitlist and have not yet shipped a unit. That's a meaningful signal that the demand is real even if the product isn't. The $165M is a bet on a category that has been promised for 40 years and has never materialized at consumer scale. The BofA forecast above suggests the window is now.

Read more - TechCrunch

Last week we covered Starship V3 and SpaceX's push toward Flight 12 in mid-March. Here's where things stand now. The launch window has slipped to NET April, according to the current schedule. The mid-March target was always soft, and a damaged earlier booster forced SpaceX to swap in Booster 19, which completed its first static fire test on Pad 2 this week. That's the critical milestone before a full stack rollout and launch attempt.

The stakes on this flight haven't changed. V3 carries over 100 tons to low Earth orbit versus about 35 tons for V2. It's the first version of Starship capable of flying to Mars and the version NASA needs for Artemis 3's lunar landing mission. Full reusability, catching the upper stage with the tower arms, is the breakthrough SpaceX is targeting this year.

If they nail it, the cost of putting things into orbit drops to under $100 per pound. A month delay is noise. The mission is still on.

Read more - Next Spaceflight | IBTimes

There's a single thread running through everything this week: the infrastructure layer for the next decade is being built right now, and the companies laying the foundation are moving faster than anyone expected.

NVIDIA's $1 trillion demand signal is not just a revenue forecast. It's a statement that every major industry, automotive, logistics, manufacturing, healthcare, has accepted that AI and physical robotics are going to be core operating infrastructure, not nice-to-haves.

When Huang announces that Uber, Nissan, BYD, and Hyundai are all building their autonomous fleets on NVIDIA's stack, that's not vaporware. Those are purchase orders. The world is being rewired and the bill is coming due.

The Bank of America humanoid forecast and Sunday's fundraise are telling the same story from a different angle. The robot revolution isn't coming because people are excited about robots. It's coming because there won't be enough workers to do the jobs that need doing.

Demographics don't care about your feelings about automation. Japan and Germany are already living the labor shortage that the U.S. will feel in the next decade, and the cost curves on humanoid hardware are following a trajectory that looks a lot like solar panels and EVs: too expensive to matter, then suddenly too cheap to ignore.

The OpenAI mini and nano launch is easy to scroll past as a developer story, but it matters for everyone. The tiered model architecture they're building is the blueprint for how AI agents will work inside every app, every workflow, every company within a few years. Most of what AI does in the background of your life in 2028 will be handled by cheap, fast subagents. The flagship models are the brain. The minis are the hands. That division of labor is being priced and shipped right now.

We're watching the infrastructure for the next 20 years get built in real time. The question worth sitting with this week: in your career and your life, where are you a builder of this infrastructure, where are you a beneficiary of it, and where might you be on the wrong side of it?

What part of this week's story surprised you most?

Confusing, jargon-packed, and time-consuming. Or quick, direct, and actually enjoyable.

Easy choice.

There’s a reason over 4 million professionals read Morning Brew instead of traditional business media. The facts hit harder, it’s built to be skimmed, and for once, business news is something you actually look forward to reading.

Try Morning Brew’s newsletter for free and realize just how good business news can be.

Check it out

The AI economy is booming, and smart entrepreneurs are already profiting. Subscribe to Mindstream and get instant access to 200+ proven strategies to monetize AI tools like ChatGPT, Midjourney, and more. From content creation to automation services, discover actionable ways to build your AI-powered income. No coding required, just practical strategies that work.

Subscribe to Get Your Free Guide