Private Markets Update & Deal Flow

Kalshi has officially banned users located in India from trading on its platform. The change was added to its June 17 member agreement, placing India on its list of restricted jurisdictions, which now runs to 55 countries and territories. Anyone domiciled, organised in, or located in India is barred from trading event contracts, a sharp reversal from Kalshi’s October 2025 announcement that it was expanding into 140-plus countries, including China and India.

The move follows months of escalating pressure. India’s Promotion and Regulation of Online Gaming Act 2025 came into force on May 1, classifying prediction markets as online money games and carrying criminal exposure, including possible imprisonment, for anyone who facilitates or finances them. MeitY blocked rival Polymarket around May 21 and warned VPN providers in April that helping users bypass the restrictions could expose them to legal consequences. Kalshi held out longer than most; its legal counsel said as recently as mid-May that the company had not received a shutdown order and would comply “should they make them.” A few weeks later, it could no longer ignore the enforcement action.

The timing matters beyond India. Nine European gambling regulators (Belgium, France, Germany, Italy, the Netherlands, Poland, Portugal, Spain, and Switzerland) signed a coordinated declaration on June 17, the same day as Kalshi’s amendment, committing to act together against unlicensed prediction markets. It is the first time European regulators have moved this way in unison, and it landed during the opening days of the World Cup, prediction markets’ biggest betting season.

India and Europe moving against prediction markets in the same week that Kalshi is raising at a USD 40B valuation is a reminder that the light-touch US regulatory environment that fuelled its growth is the exception internationally, not the rule. Expect the eventual IPO prospectus to spend real space on jurisdictional risk.

Here is a trade with no money down. In early 2025, Donald Trump Jr. joined prediction-market platform Kalshi as a strategic adviser and was handed roughly USD 300k in equity, not a cent of his own committed. Eighteen months later, that paper looks inspired: Kalshi’s valuation has gone from under USD 2B to USD 22B in its May 2026 Series F, with talks now underway at around USD 40B. Even after dilution, the freebie is worth a fortune.

The timing tells the story. Kalshi has sprinted under a Washington that stopped swinging at it (the CFTC dropped its appeal in 2025 and closed its Polymarket probe) and has since pushed into crypto perpetual futures. Both Kalshi and Polymarket have been on a tear since calling Trump’s 2024 win, now clearing billions in monthly volume, lawsuits and state pushback notwithstanding.

What the street is really chewing on is the pattern. The Trump family has woven financial ties into crypto, drones, online firearms, and sports, all sitting downstream of current policy, and the Kalshi stake adds another line to a ledger drawing fresh conflict-of-interest scrutiny. The bigger question is what the eventual IPO disclosures reveal. Until then, this stays the rare position that costs nothing to hold and gains value every quarter, which is exactly why everyone is watching it.

A free stake that has gone from under USD 2B to USD 22B, with USD 40B talks already underway, is no longer a footnote. It is a live conflict-of-interest story that will follow Kalshi straight into its IPO disclosures, and one regulators reviewing the listing will not be able to ignore.

Anthropic is rapidly expanding its AI infrastructure in the Asia-Pacific region, with 8 of its 13 open compute roles based in Australia and Japan. The hiring spree signals that both countries are becoming key locations for the company’s next generation of data centres as demand for its models continues to surge. Anthropic has admitted that its rapid growth is placing “an inevitable strain” on infrastructure and reliability.

The expansion follows explosive growth. Anthropic recently raised USD 65B at a USD 965B valuation, while its annualised revenue run rate jumped from around USD 9B at the end of 2025 to over USD 47B in May 2026. Australia is emerging as an attractive hub thanks to abundant land, renewable energy, political stability, and Five Eyes security ties to the US, though strict copyright laws and uncertainty around AI training data remain obstacles. Japan offers a reliable power grid, advanced internet infrastructure, and government support, with Microsoft already committing USD 10B and GMI Cloud USD 12B to sovereign AI projects there.

Despite the momentum, the binding constraint is increasingly power, not land or financing. Anthropic says it is focusing expansion on democratic countries with secure supply chains, but across Asia-Pacific, electricity access is becoming the defining constraint on how fast AI infrastructure can actually scale.

Anthropic is not short of capital or demand; it is short of power. The companies that win the next phase of AI infrastructure will be the ones that can secure grid capacity in democratic, low-risk jurisdictions, not the ones with the deepest pockets. Australia and Japan are positioning to be exactly that.

OpenAI made its first-ever appearance at Cannes Lions this week, and it was not there for the rosé. The company pitched two things to marketers: its fast-growing ChatGPT ad business and Codex, its AI coding tool, while quietly setting the table for a possible IPO later this year. Chief Revenue Officer Denise Dresser put it plainly: “We are clearly in the advertising business now.”

The number OpenAI is whispering to ad executives is eye-watering: a USD 100B annual ad opportunity by 2030, roughly half of Meta’s current ad revenue. Ads currently appear only for Free and Go-tier users, clearly labelled and, OpenAI insists, never influencing the chatbot’s answers. The company says roughly 20% of ChatGPT queries carry direct commercial intent, with travel, retail, beauty, and financial services pulling the strongest early results, and projects USD 2.5B in ad revenue this year alone, already past USD 100M in annualised revenue within the first six weeks of launch.

The Codex angle was the more interesting tell. OpenAI demoed brands and agencies spinning up marketing apps and campaign tools from plain-language prompts, with Codex now counting more than 5M weekly users, up roughly 400% this year. Executives were careful to frame it as augmenting agencies rather than replacing them. The agencies in the room may want that in writing.

OpenAI is doing to Google what Google once did to print advertising: meeting intent at the exact moment it appears. The bet is that AI-native search becomes the next great ad channel before the IPO roadshow starts. Whether it can build that machine fast enough without spooking users who came for the AI, not the commercials, is the open question.

China’s humanoid robot boom is accelerating faster than expected. Morgan Stanley has doubled its 2026 shipment forecast for the second time this year, now expecting 50,000 humanoid robots to ship in China, up from an initial 14,000 in January and a March revision of 28,000.

The industry is moving rapidly from demonstrations to real-world deployment. Chinese robots are already working in factories, logistics centres, restaurants, and retail, backed by strong government incentives and a deepening domestic supply chain. State Grid Corporation placed a roughly USD 1.0B order for 500 humanoid and thousands of quadruped robots in the first half of 2026 alone, while SF Express and China Post have begun deploying humanoids for sorting and handling at logistics centres.

Morgan Stanley now estimates China’s humanoid robot market will reach USD 2B this year and expand to USD 15B by 2030, with annual shipments climbing to 446,000 units. Chinese firms shipped more than 80% of the world’s humanoids last year and took the top five spots by volume; Tesla does not plan public Optimus sales until the end of 2027. Geopolitics remains the biggest risk as Chinese makers expand into global markets, but the commercialisation lead is widening, not narrowing.

The second upward revision in a single year is the signal, not the number itself. A market that outpaces professional forecasts this consistently has entered a different growth phase than a normal tech cycle. Morgan Stanley’s own preferred play is not a robot maker at all but Leaderdrive, a Shanghai-listed precision-component supplier, the classic picks-and-shovels bet on a market nobody can yet call a winner in.

The honeymoon is over, but the rocket is still well above the launch pad. SpaceX shares fell 16.4% on Monday to close at USD 154.60, the stock’s worst single day since its IPO, capping a three-day slide of roughly 23 to 27% from its USD 226 peak and erasing more than USD 600B in market value. Shares remain about 14 to 22% above the USD 135 IPO price, so everyone who got in at the largest listing in history is still in the green. After a debut that briefly vaulted SpaceX past Amazon and Microsoft, a breather was arguably overdue.

What broke the spell was not company-specific. Bond yields jumped as the street priced in possible Fed hikes tied to inflation pressure from the Iran conflict, and richly valued growth names took the hit; SpaceX just happens to be the biggest and newest one in the window. The company also confirmed its first-ever bond offering, targeting roughly USD 20B in senior unsecured notes to refinance the bridge loan it used to acquire xAI in February, revealing that a meaningful slice of its USD 86.2B IPO haul was already earmarked for debt. The first week of SPCX options trading, which let short sellers finally hedge, amplified the move.

There is real machinery underneath the volatility. SpaceX disclosed USD 100.8B in cash and cash equivalents as of June 19, and the Cursor acquisition's AI contracts (Anthropic and Google among them) are recurring revenue landing outside the core rocket-and-satellite business. But xAI posted a USD 6.4B operating loss in 2025, all 11 of its original co-founders have since departed, and Grok still trails ChatGPT, Gemini, and Claude on adoption. The bigger question is not whether SpaceX stumbled this week, but whether a company this early in its public life can convert launch-day hype into the durable AI franchise its valuation is betting on.

A USD 2T-plus valuation trading at over 100x trailing revenue was always going to be volatile on its way to finding a real price. The selloff is the market doing its job, not a verdict on the company. What matters more than this week's move is what is still coming on the lock-up calendar below.

At SpaceX’s IPO, only about 4.9% of shares were freely tradable, an exceptionally tight float that did most of the work behind the stock's initial run to USD 226. Over the following months, additional shares become eligible for sale through a structured lock-up release schedule: 11.8% by August 8, 15.2% by August 20, 17.7% by September 9, climbing toward 40% by December 8, 2026.

The most important date sits a year out. On June 12, 2027, Elon Musk’s 46.1% stake exits its lock-up entirely. In a single day, the theoretical free float jumps from around 50.8% to 96.9%, the largest single increase in potential share supply this market is likely to see for a long time. The process is expected to be fully complete by September 2027.

Lock-up expirations do not guarantee anyone sells, but the sudden availability of that much supply is a structural risk that has nothing to do with SpaceX's fundamentals. Historical lock-up studies find a modest average abnormal return of around negative 1.5% over a few days, with a sharp jump in volume, though outcomes vary widely by name. Investors holding SPCX through 2026 and into mid-2027 will be watching insider selling activity at every one of these dates, particularly the one a year from now.

This is the chart every SpaceX holder should have pinned above their desk. The June 2027 unlock is not a footnote buried in the S-1; it is the single largest supply event scheduled into any major US stock over the next two years. Whatever SpaceX's AI and rocket businesses are worth by then, the share price will have to absorb that supply shock regardless.

SpaceX just handed a little-known Israeli chip startup one of the most coveted endorsements in tech. Xsight Labs landed SpaceX, which picked its X2 programmable Ethernet switch as the networking backbone of Starlink’s next-generation V3 satellites. Each V3 satellite is designed to deliver more than 1 terabit per second of fronthaul throughput, over 10x the capacity of the current V2 Mini satellites, and the X2 delivers 12.8 Tbps of switching capacity at under 200W on TSMC's N5 process.

Satellites are brutally power-constrained, and Xsight’s low-power, fully programmable architecture is exactly what wins when every watt counts. SpaceX did not buy the brand name; it bought the engineering, after the chip passed a full suite of environmental qualification tests for vibration, radiation, and thermal stress.

Now it gets interesting. Founder and chairman Avigdor Willenz has done this before: his previous semiconductor companies have gone to Marvell, Amazon, and Intel. The man builds chip businesses that strategic buyers cannot resist. With SpaceX's name now attached, Xsight looks less like a supplier and more like a target, and Broadcom, Nvidia, and AMD are all hungry for differentiated networking silicon. SpaceX's supplier ecosystem is quietly becoming a scouting ground for the next AI and space-infrastructure winners, long before they hit public markets.

Willenz's pattern is consistent enough to set a watch on: build the differentiated chip, land the marquee customer, sell to the strategic buyer who can't risk a competitor getting there first. SpaceX's endorsement just started that clock again, and Broadcom, Nvidia, and AMD all know it.

Here is a trade aging nicely. Peter Thiel’s Founders Fund planted its flag in a Berlin drone shop before Europe decided it urgently needed one, and this week the rest of the market scrambled to catch up. Stark Defence just pulled in EUR 500M at a EUR 3.2B valuation, with Founders Fund and Sequoia among the names writing checks, alongside the NATO Innovation Fund, Döpfner Capital, Air Street Capital, 201 Ventures, and Project A.

Its signature product is the Virtus, a loitering munition built to take out targets by flying into them, and this is not a deck-and-a-dream story. Stark already holds a roughly EUR 269M contract to supply drones to Germany's 45th Armoured Brigade stationed in Lithuania, alongside fellow startup Helsing. That is revenue, not just narrative.

Yes, Thiel backing a firm Berlin leans on has made some politicians twitchy. But for the Founders Fund side, the structure is close to ideal: Thiel holds under 10%, diluted further by this round, the company is European-run, and no single shareholder gets near the sensitive technology. Upside exposure to a national-security darling without the cap-table baggage that spooks governments. And the money keeps getting louder, piling into a tight cluster (Stark, Helsing, Quantum Systems) each a bet on Europe's eventual defense champion. Founders Fund did not just buy a drone maker; it bought an early, well-governed seat at a rearming continent.

The only question left is how big the prize gets. With Sequoia, the NATO Innovation Fund, and a cluster of European defense champions all converging on the same handful of names, Founders Fund's early, well-governed position in Stark looks less like a contrarian bet and more like the consensus trade arriving a year late.