Amazon built a robot that can feel.
Vulcan is its first robot with a sense of touch, built to pick and stow items in those cramped storage pods without crushing them.
- Force-feedback sensors: it feels contact and how hard it's pushing, so it won't damage your order
- Handles ~75% of item types, at near-human speed
- Knows when to tap in a human for the rest
- Working in Spokane + Hamburg, rolling out across the US and Europe
Robots learned to see years ago. Touch is the real unlock for physical AI.
Masayoshi Son (CEO of SoftBank) to CNBC: the next trillion-dollar company will come from physical AI and robotics. Humanoid and industrial.
This is the guy who bet early on Alibaba, ARM and OpenAI.
His framing:
- The stack is "more than 10x, probably 50x bigger than dot-com." The biggest tech revolution mankind has experienced, and just the beginning.
- It could run 50 to 100 years.
- On a crash? "There may be some correction, but that will be the best investment opportunity." He points to electronics and autos after the 1929 crash: they fell, then climbed for the next century.
And he isn't just talking. SoftBank just committed $87B to AI infrastructure in France, and bought ABB's robotics arm for $5.4B.
The bet is simple: the next Alibaba is a robot company.
Today's AI map: mostly red but two segments were untouched.
Quantum -4 to -10%, optical pure-plays down hard, robotics dumped, even NVDA red.
Two segments stayed green across the board: memory and semi equipment.
Why? They're the least speculative AI trade. Memory's in its worst shortage in ~15 years (DRAM +~60% this quarter, NAND +~75%, the big makers sold out through 2026). Equipment is what everyone buys to make more.
When the froth comes off, money runs to scarcity and backlog. You don't guess who wins. You bet chips stay scarce.
AAOI ripped ~28% in two days. Quick primer on the optical transceiver trade:
These companies make the parts that turn data into light, to move it between AI chips and across data centers. Pure AI picks-and-shovels.
- $AAOI: small, explosive pure-play. +340% YTD, $324M+ in 800G/1.6T orders, loss-making, now has a 2x leveraged ETF. Swings 2-3x harder than peers.
- $COHR $LITE $FN $CRDO: the bigger, steadier names. NVIDIA put $2B into Lumentum and signed an optical deal with Coherent.
Last week the group dipped, then ripped together on Jun 2.
AI's data has to move as light. These make it move.
You know what they say...
No risk, No hanging out with lions in the savannah, dressed in white linen, driving an old jeep with cooling problems, knowing you're gonna be rich because you're positioned.
Ray Dalio went on Bloomberg today and said the AI boom is showing signs of a bubble that will eventually burst, a lot like the 2000 dot-com crash.
But his point wasn't 'sell.' It was this:
'All great technology changes produce bubbles.' Nobody can time it perfectly. So every company faces one choice: spend a fortune to win the new market and risk overspending, or hold back and lose the market entirely.
History backs him up.
The ones who bet big won:
- Amazon poured billions into the cloud before anyone asked
- Apple killed its own iPod to build the iPhone
- Nvidia bet on GPUs years before AI existed
- Netflix torched its DVD business for streaming
The ones who protected the past lost:
- Kodak invented the digital camera in 1975, then buried it. Bankrupt 2012.
- Blockbuster passed on buying Netflix for $50M. Bankrupt 2010.
- Nokia owned 40% of phones and shrugged at the touchscreen. Gone.
- Intel passed on the first iPhone chip. Missed mobile entirely.
That's the AI capex debate in one picture. The spending looks insane. It might pop. But the fatal move, every single time, was sitting it out.
The bubble bursts. The technology doesn't.
Deeper Dive into MRVL: An AI data center isn't one giant chip. It's thousands of NVIDIA GPUs that only work if they can talk to each other instantly. That 'talking' is the real bottleneck. And it's mostly Marvell's job.
What Marvell makes:
- Optical chips that move data between GPUs as light, because past ~10 meters copper can't keep up. Latest speed: 1.6 terabits per second.
- The custom AI chips Amazon, Google and Microsoft design in-house (Trainium, Maia, TPU), built on Marvell's platform. A 10B+ business by 2029.
- The switches and high-speed wiring that route all of it.
That's why NVIDIA put $2B into Marvell, and why Jensen Huang called it 'essential' on stage.
The GPU is the muscle. Marvell is the nervous system.
SpaceX going public is around the corner. Possibly the biggest IPO ever.
Ticker SPCX, target valuation ~$1.75 trillion.
The twist: it's barely a rocket company.
· ~70% of revenue is Starlink (satellite internet, 10M+ subscribers)
· it now owns xAI / Grok, so it's an AI company too
· it wants data centers in space
Reality check: ~$5B loss in 2025 on ~$18.6B revenue.
You're not buying a trip to Mars. You're buying internet and AI that happens to launch rockets. SPCX
Records across the board today. The S&P cleared 7,600 for the first time by rising a heroic 0.13%.
$MRVL +33%, biggest day ever, because Jensen Huang called it the next trillion-dollar company on stage. Praise is a catalyst now.
$GOOGL -4% raising $80B for AI ($10B from Berkshire) the same week Buffett warned about 'gambling.'
Bitcoin -6%. $HPE +19% on earnings.
3 names = 40% of this year's S&P earnings revisions. But sure, broad rally.
RoboStrategy ($BOT) is a new Nasdaq fund built around one idea: giving regular investors a way into private robotics.
The biggest names in humanoid robots Figure, Apptronik, Standard Bots and others are private, normally only accessible to venture capital. $BOT holds stakes in a basket of them, so that exposure now sits behind a ticker anyone can buy. It listed in May 2026 and is run by @Rewkang (Mechanism Capital).
A few things worth understanding about how it works:
· It's a closed-end fund, so its share price can trade above or below the value of what it holds.
· Its holdings are private companies, valued by periodic estimates rather than a live market price.
· The fund can issue new shares to raise capital.
· Since listing, it has traded as high as ~$59 and recently around ~$24.
It's a new way to get exposure to a part of the market that used to be closed off. As with any closed-end fund, the price and the value of what's inside aren't always the same worth understanding before you decide either way. $BOT
You all heard him, tell your friends about real utility built on pump fun.
Live now:
- Daily research cards on the AI buildout: chips, robotics, quantum, datacenters, power.
- The trade under the headline. Everyone sees the news. We show you what catches the bid.
- Fund-flow tracking. KOLscan, but for investment funds and these KOLs are legally required to post their bags every quarter.
- Supply-chain map. The full AI stack, foundry to power. See the chokepoint, find the value.
In development:
- Conviction scores. A heatmap of where capital's rotating. Ranked.
- Agentic council. AI agents running the research grind at scale.
Bluechip is the next Bluechip.
Quantum computers are powerful but fragile their building blocks forget everything in millionths of a second. That's the main thing stopping them from being useful.
Microsoft's new chip (Majorana 2) holds on about 1,000× longer ~20 seconds.
Think: a phone battery that lasts a lifetime instead of dying in a blink.
Big if true $MSFT