TechWithAdit (TWA)

Meta just launched AI glasses at $299 — its most aggressive push yet to own the post-smartphone era. As the battle for the next computing platform heats up between Meta, Google, Snap, and Apple, Masayoshi Son is speaking to SoftBank shareholders right now in Tokyo, tying his entire company's net asset value to a 10-year bet on Super AI arriving within our lifetime.

On the ground, Nvidia chips are selling for $1.1M on China's black market as the US-China compute war plays out in real time. Infosys chairman Nandan Nilekani told shareholders AI will amplify, not replace, his company. And Claude Code creator Boris Cherny is now admitting that once AI writes 100% of your code, the bottleneck shifts from engineering to ideas — and most companies are nowhere near ready for that conversation.

Meta and EssilorLuxottica have launched a new range of AI smart glasses starting at $299 — the first Meta glasses sold under its own brand rather than Ray-Ban or Oakley, and far cheaper than last year's $800 Ray-Ban Display. The glasses are screen-free, built around audio, camera, and voice AI powered by Meta AI's Muse Spark — the first model from Meta's Superintelligence Labs. New designs include rectangle frames and a slim oval collection co-created with Kylie Jenner. Available immediately in the US, UK, Canada, Australia, and Europe.

The competitive context sharpens the launch. Snap's AR Specs launched last week at $2,195. Google and Samsung are collaborating on AI glasses later this year. Apple is in development. Meta currently holds 69.2% of the smart glasses market with shipments up 167% in Q1 2026 year-on-year. Zuckerberg said daily active users of its glasses have tripled year-over-year. The $299 price is a deliberate land grab — get the hardware in as many hands as possible before rivals reach mass market.

The insights: The smartphone had a price point that made it universal. Meta is trying to do the same for AI wearables — undercut everyone, dominate distribution, then layer in services. At $299, this is no longer a niche product. It is a consumer electronics play, and every company in the space now has to respond to that number.

The Financial Times is reporting that Nvidia's DGX B300 server is now hitting $1.1 million on China's unofficial market — more than double its official US price — as export controls have created severe scarcity in the world's second-largest AI economy. Despite the ban, chips are clearly flowing through third-party channels, grey-market distributors, and intermediary countries. The price premium China is willingly paying is itself a signal of how critical Nvidia's compute is considered for staying competitive in the AI race.

The story reveals a contradiction at the heart of US export policy. Controls are working well enough to create significant scarcity and cost pain for Chinese AI labs. They are not working well enough to cut off access entirely. Chinese companies are paying more, waiting longer, and sourcing through riskier channels — but they are still getting the chips. Meanwhile, this is precisely the pressure driving China's $295B national AI grid mandate requiring 80% domestic silicon.

The insights: For any operator building AI infrastructure globally, this story reframes the chip supply question. The US-China AI compute split is not theoretical anymore — it is priced into black markets. The two-stack AI world Son is building SoftBank around is already forming at the hardware layer.

Speaking at Infosys' 45th Annual General Meeting yesterday, chairman Nandan Nilekani pushed back directly against growing investor concern that AI threatens the IT services model. "AI will not replace a company like ours. It will amplify those who move with purpose and adapt with speed," he told shareholders. His argument: software development requires organisational context, systems integration, cybersecurity, testing, governance, and architectural knowledge that goes far beyond writing code — areas where human-AI collaboration beats pure automation. Infosys is already working with 90% of its top 200 clients on AI initiatives and sees a $400 billion opportunity in AI-first services by 2030.

Infosys reported fiscal 2026 revenue of $20.2 billion, with 3.1% constant-currency growth and $3.7 billion in free cash flow. Nilekani framed the deployment gap in large enterprise clients as the immediate business opportunity: most organisations have bought the AI vision but have not closed the gap between pilot and production.

The insights: The fastest-growing AI companies and the largest IT services firms are converging on the same client base. Nilekani's argument — that closing the enterprise deployment gap is where $400 billion lives — is the same bet Infosys, TCS, and Wipro are all making. The question is whether Indian IT can move fast enough to own that gap before hyperscaler professional services teams do.

Superhuman Acquires GPTZero — AI Detection Goes Enterprise Mainstream

Superhuman, formerly Grammarly with 40 million users, has acquired GPTZero — the Princeton-born AI detection startup with 19 million users and $30M ARR. Founded in 2023 by Edward Tian and Alex Cui, GPTZero detects AI-written text, hallucinations, fake citations, and AI-generated images in real time. All 30 employees and both co-founders join Superhuman. The technology integrates into Superhuman Go, which works across one million apps. The irony is also the insight: the same platform that helps people write with AI just acquired the tool that detects when content was written by AI.

The insights: Content authenticity is no longer academic — it is a legal and brand requirement for publishers, enterprises, and regulated industries. By acquiring GPTZero rather than building in-house, Superhuman gets three years of battle-tested models and 19 million users in one move.

Masayoshi Son is speaking live at SoftBank Group's 46th Annual General Meeting in Tokyo right now, and the WSJ is reporting he is setting the most ambitious net asset value target in the company's history — tied directly to his timeline for Super AI. The thesis is unambiguous: Artificial Super Intelligence, AI 10,000 times smarter than any human, will arrive within 10 years. SoftBank's entire portfolio is structured to be the dominant infrastructure platform when it does.

The numbers behind that declaration are staggering. SoftBank's NAV has climbed to approximately ¥48 trillion, up from ¥25.7 trillion just a year ago. Arm Holdings, which SoftBank owns ~90% of, accounts for more than 50% of that NAV. OpenAI, where SoftBank's cumulative investment is heading toward $64.6 billion by October 2026, accounts for over 20% — with OpenAI's valuation surging from $260 billion to $730 billion within the year. SoftBank has overtaken Toyota as Japan's most valuable company. Today, Son is formally amending SoftBank's articles of incorporation to add AI, semiconductors, robotics, and data centers as core business purposes.

"I think this is more than 10x, probably 50x bigger than dot-com," Son said at his June Paris briefing. The NAV target he is setting today ties that conviction to a specific financial commitment, made in front of shareholders, on the record.

The insights: Son has been called a visionary and a reckless gambler in the same sentence. He lost $70 billion in the dot-com crash and came back richer. He nearly destroyed his credibility on WeWork and rebuilt it through Arm. Today's AGM is his most audacious declaration yet — and unlike his earlier bets, the underlying technology is already delivering returns. The question for every operator is not whether his ASI timeline is correct. It is this: if the world's most aggressive AI investor is publicly tying his company's financial targets to a 10-year Super AI timeline, what are you building yours around?

Was this email forwarded to you? Don't miss any updates — Subscribe to TechWithAdit for sharp, no-noise tech intelligence. Stay sharp. — Adit