Welcome, AI Insiders!
If you plan or pay for AI, read this. Huang is calling for three to four trillion in new infrastructure by decade end. On the ground, teams adopt AI that assists not replaces. At home, Copilot moves to the biggest screen. In health, MIT says we can pick flu shots with foresight, not luck.
📌 In today’s Generative AI Newsletter:
NVIDIA softer guide, but $3T–$4T AI capex vision
Netstock “assist, don’t replace” wins skeptical staff
Copilot × Samsung on-screen assistant for living rooms
MIT VaxSeer predicts dominant flu strains ahead of season
🟢 NVIDIA’s AI Boom Still Has Room to Run, Says Jensen Huang
Nvidia CEO Jensen Huang in Washington D.C., U.S., July 23, 2025. REUTERS/Kent
NVIDIA CEO Jensen Huang pushed back on talk of an AI slowdown after the chipmaker’s latest forecast fell short of Wall Street’s sky-high expectations. Speaking after earnings, Huang predicted $3T–$4T in AI infrastructure spending by decade’s end, calling the moment “a new industrial revolution.”
The details:
Sales outlook: NVIDIA’s Q3 revenue guidance came in softer than expected, partly because China sales were excluded due to trade tensions. Shares slipped nearly 2% to $178.43.
China factor: NVIDIA struck a deal with Trump to export H20 chips in exchange for giving the U.S. government a 15% cut of China sales. Huang even suggested he would extend the offer to Blackwell chips.
Massive demand: Hyperscalers like Microsoft and Amazon are expected to spend $600B on data centers this year. Huang said NVIDIA captures about $35B for each $60B facility.
Everything spoken for: NVIDIA new Blackwell chips are already booked through 2026, and its older Hopper processors are still in demand.
NVIDIA’s stock has been the flagship of the AI trade, but signs of fatigue are creeping in. Huang insists we are still early, pointing to mega-cap capex and sold-out chips as proof that the AI boom is far from peaking.
🤝 How One Company Is Easing Small Businesses Into AI
Image Credits:Jetta Productions / Getty Images
Netstock isn’t new. The inventory management company has been around since 2009. What is new is its Opportunity Engine, a generative AI tool that plugs into ERP systems and spits out real-time stock recommendations. Since launch, it’s served up more than 1M suggestions, with three-quarters of customers seeing tips valued at $50K+.
What’s actually happening on the ground:
Family business test case: Bargreen Ellingson, a 65-year-old restaurant supply company, was skeptical. Its staff only warmed up after the AI was pitched as optional, not a takeover tool.
AI as a crutch, not a boss: Warehouse workers use the system to catch errors and sift through endless reports faster, but final calls still come from humans.
Skill shift: Even junior staff can spot useful insights in seconds. But long-term, it could reduce the demand for data science expertise inside companies.
Built-in restraint: Netstock keeps recommendations in the dashboard where they can be accepted or ignored. No chatbots, no “black box” conversations.
Netstock’s pitch is deliberately boring and that might be why it works. The Opportunity Engine isn’t revolutionizing supply chains, but it reshaping day-to-day warehouse work. The real story is not about efficiency gains but about how much decision-making companies are willing to hand over before the balance tips.
📺 Microsoft Copilot Lands on Samsung TVs and Monitors
Image source: Microsoft
Copilot is moving from laptop screens to living rooms. Microsoft just announced that its AI assistant will ship on Samsung’s 2025 TVs and smart monitors, giving users an animated, blob-like character that talks back, lip-syncs, and fields everything from binge recommendations to weekend planning.
Here’s what’s rolling out:
On-screen AI: Copilot shows up as a blob character that reacts in real time with expressions and voice, turning your TV into a more interactive presence.
Integrated into Tizen OS: Users can launch Copilot from Samsung Daily+ or by pressing the mic button on their remote, no extra install required.
Spoiler-free recaps and group picks: Whether you left The Crown mid-season or need a movie everyone in the room will tolerate, Copilot will surface suggestions.
Beyond TV: Ask about the weather, local plans, or even for a pick-me-up after a rough day. Logged-in users get personalization, memory, and preference recall.
TVs have been called smart for years, but this pushes them into something new: social. Copilot now shows up as a visible companion on the biggest screen at home, ready to interact with everyone in the room, placing itself at the center of family time.
🦠 MIT Builds AI That Predicts the Flu Before It Mutates
Image: Alex Gagne
Flu vaccines are picked months before the season begins, a gamble that can leave millions exposed if the wrong strains take hold. MIT researchers have introduced VaxSeer, an AI tool trained on decades of viral data that uses deep learning to predict which strains will dominate and which vaccines will hold the line.
The details:
Two prediction engines: One forecasts how flu strains spread, the other estimates how well a vaccine will neutralize them.
Proven record: In a 10-year study, VaxSeer outperformed the WHO’s choices in nine of ten H3N2 seasons and matched or beat them in most H1N1 seasons.
Ahead of the curve: In 2016, it identified a strain that global health authorities did not select until the following year.
Inside the system: The tool combines protein language models, differential equations, and lab-based measures of vaccine effectiveness.
Bigger horizon: The same approach could be applied to fast-evolving threats like antibiotic resistance and drug-resistant cancers.
Flu mutates constantly, and the vaccines that protect us often lag behind. VaxSeer offers a way to anticipate viral change rather than chase it, giving health officials a sharper edge in the fight to keep seasonal outbreaks under control.
🚀 Boost your business with us. Advertise where 13M+ AI leaders engage!
🌟 Sign up for the first AI Hub in the world.
