SpaceX files for the largest IPO in history. The US bets $2B on quantum.

SpaceX filed its S-1 on May 20, targeting a $1.75T valuation, a $75B raise, a June 12 listing on Nasdaq under SPCX, with Goldman Sachs leading 21 underwriters and an unusual 30% retail allocation of the float.

The S-1 reveals $18.7B in 2025 revenue (33% YoY growth), with Starlink generating $11.4B — the company's only profitable division, subsidising a $4.94B GAAP net loss driven by $10.1B in Q1 2026 capex, $7.7B of which went to AI infrastructure (Colossus, the xAI compute cluster).

On May 27, the Space Force awarded SpaceX a $2.29B Other Transaction Authority contract for the Space Data Network Backbone — a LEO constellation designed to provide secure, high-bandwidth communications for Golden Dome missile defence and broader US military operations.

The SpaceNews thesis Marc shared is live: space now has recurring revenue infrastructure, settled customers, and a daily launch cadence. The Netscape moment — when a browser made the internet usable, triggering institutional investment — has its space equivalent in Starlink crossing 10 million paying subscribers and SpaceX listing on Nasdaq in the same quarter.

The S-1 forces a sector-wide repricing: every private space company will now be benchmarked against a public SpaceX multiple. Companies with proximity to Starlink's commercial data layer (ground stations, spectrum holders, earth observation) benefit from the rerating; those competing directly face an elevated comparison point.

The $2.29B Space Force contract is strategically timed: it demonstrates institutional government revenue beyond Starlink before the roadshow begins, directly addressing the S-1's core risk — that the valuation depends too heavily on a single business line.

SpaceX pre-IPO access remains limited to qualified institutional buyers through the roadshow; the June 12 listing will be the first public entry point at scale. The 30% retail allocation is unusually large for a deal of this size — a deliberate move to broaden the shareholder base and create secondary market depth.

The S-1 also discloses a $1.25B per month Anthropic compute deal (up to $45B over 3 years) — meaning SpaceX is simultaneously a launch company, an ISP, an AI infrastructure provider, and a military contractor. The correct comparable is not Boeing or Airbus; it is a combination of AT&T, AWS, Lockheed Martin, and Nvidia.

Risk caveat: SpaceX carries $60.5B in total liabilities, a $41.3B accumulated deficit, and Musk holds 85.1% of voting control through a dual-class structure. Investors buying SPCX are financing Musk's priorities, not directing them — the same governance structure that defines Tesla and xAI.

The Commerce Department signed Letters of Intent with 9 companies for a combined $2B under the CHIPS Act, with the government taking equity stakes — named recipients include Quantinuum ($100M), Infleqtion ($100M), IBM, IonQ, Rigetti, D-Wave, and Pasqal, with additional companies expected to be confirmed.

On May 27, Quantinuum filed its IPO prospectus, targeting a ~$12.7B valuation on Nasdaq — backed by Honeywell (majority shareholder), with a commercialisation roadmap targeting its first fault-tolerant Apollo system in 2029 and a revenue trajectory that has grown significantly from its trapped-ion hardware base.

The two events are structurally linked: the LOIs require recipients to demonstrate technical milestones against benchmarks set by DOE's Office of Science — effectively a government-run performance-based funding mechanism that narrows the field while accelerating the surviving companies.

The government equity stake model changes the investment dynamic fundamentally. In the 2022 CHIPS Act, federal money rebuilt domestic semiconductor manufacturing capacity — private capital followed. Quantum is now being treated the same way: the US is picking infrastructure-layer winners and taking upside, not just funding research.

Quantinuum's IPO filing creates a public benchmark valuation for the first time. At $12.7B, it prices in a 2029 fault-tolerant system that hasn't been built yet — but it also creates a reference point for every private quantum company (IonQ's public peers, PsiQuantum's next round, Pasqal's SPAC filing) to be measured against.

The coincidence of government investment and IPO activity in the same week mirrors the semiconductor moment of 2022–2023: when public policy and public markets aligned, private capital flooded in. Quantum may be at that same inflection.

The nine LOI recipients are a government-curated shortlist of quantum companies most likely to achieve utility-scale systems — an extraordinary de-risking signal for LPs evaluating quantum positions. Appearing on that list is now itself a commercial credential.

The Quantinuum IPO will be the first pure-play quantum stock accessible to public investors at institutional scale. Its ion-trap architecture, Honeywell manufacturing base, and commercial cryptography revenue (already generating income from quantum-secured keys) make it the most commercially mature quantum company approaching public markets.

Risk caveat: fault-tolerant quantum computing remains undemonstrated at commercial scale. Quantinuum's $12.7B valuation prices in a 2029 milestone that requires sustained technical progress on error correction — the same challenge that has pushed "5 years away" estimates forward for a decade.

Thea Energy raised $100M Series B led by US Innovative Technology Fund, with GIC Partners and Prelude Ventures participating — placing it among the top-funded US fusion startups alongside Commonwealth Fusion Systems and Helion Energy.

The raise was accompanied by DOE certification of Thea's preconceptual design milestone under its Milestone-Based Fusion Development programme — a formal government validation that the engineering design has cleared the first of several federally benchmarked technical gates.

Thea's stellarator architecture uses precisely shaped external magnets to confine plasma without requiring the plasma to carry its own stabilising current — eliminating the disruption risk inherent in tokamaks while introducing manufacturing complexity that modern computational design has largely addressed.

Three meaningfully different fusion architectures are now simultaneously advancing through capital raises and DOE milestones in 2026: CFS (tokamak), Helion (field-reversed configuration), and now Thea (stellarator). The sector is not converging on one approach — it is running three engineering experiments in parallel, which increases the probability that at least one delivers commercial fusion power in the 2030s.

DOE design milestone certification is more than a press release — it is a formal checkpoint in a programme where government funding is tied to technical performance. Clearing it means Thea has satisfied independent engineering reviewers that its power plant concept is internally consistent and physically plausible, a higher bar than a corporate announcement.

The US Innovative Technology Fund is Thomas Tull's deep-tech vehicle — the same investor profile behind frontier bets on nuclear, defence technology, and advanced manufacturing. Its presence signals that fusion is attracting capital from outside the traditional cleantech and energy transition pools.

Thea's stellarator approach has a specific advantage for grid operators: no plasma disruptions means no sudden power loss events, which is one of the reasons utilities and industrial offtakers have historically been reluctant to contract with fusion companies for firm power delivery.

The enabling technology stack includes precision stellarator magnet design software, superconducting coil manufacturing, and high-temperature plasma diagnostics — each a distinct investable layer that benefits from multiple stellarator programmes advancing simultaneously.

Risk caveat: stellarators have historically required extremely precise magnet fabrication to maintain plasma confinement — tolerances measured in fractions of a millimetre across metre-scale structures. Thea's thesis is that modern manufacturing and computational design have solved this problem. The prototype results will either confirm or challenge that thesis.

Hyundai Motor Group committed to deploying 25,000+ Atlas robots across its own US plants — first at HMGMA Savannah in 2028 (parts sequencing, then assembly), then Kia Georgia in 2029 — using 83% of the group's planned 30,000-unit annual production run as internal deployment before any external customer deliveries begin.

The group also announced a US actuator production facility (operated by Hyundai Mobis) targeting 350,000 units annually from 2028 — each Atlas requires 10+ actuators, making the plant sufficient to equip ~35,000 robots per year and establishing Hyundai as the only humanoid programme with vertically integrated actuator supply at scale.

Current Atlas unit cost is approximately $140,000; the group projects this falls to $30,000 when production exceeds 50,000 units annually — roughly equivalent to one US factory worker's annual wage, which is the economic threshold at which broad deployment becomes financially rational for industrial customers.

This is the largest humanoid robot deployment commitment by a single organisation globally. Hyundai is not ordering robots from a third party — it is building the factory, the supply chain, and the robots itself, and deploying them into its own highest-volume production environment. The risk profile is entirely different from a startup pilot.

The $140,000 → $30,000 cost curve is the most important number in humanoid robotics right now. At $140,000, Atlas competes with premium automation equipment. At $30,000, it competes with direct labour — and at that price point, the addressable market expands from frontier manufacturing experiments to every production line in the world. Hyundai's volume commitment is what unlocks that curve.

Barclays Research — whose report Marc shared this week — identified actuators as ~50–60% of humanoid production costs and Europe's precision engineering base as a structural advantage in that supply chain. Hyundai's US actuator plant directly responds to this: building domestic supply chain as a hedge against both Chinese component dependencies and tariff exposure, while simultaneously proving the economics of vertical integration.

Hyundai Motor (005380 KS) and Kia (000270 KS) are publicly listed — the JPMorgan presentation makes the humanoid deployment timeline concrete for equity investors pricing physical AI optionality into automotive OEM valuations for the first time.

Boston Dynamics remains privately held within Hyundai Motor Group. The 30,000-unit annual production target and US actuator plant announcement are the clearest signals yet that a Boston Dynamics IPO or partial divestiture — at a valuation reflecting robotics infrastructure, not just R&D — is a plausible near-term event.

Risk caveat: the Korean Metal Workers' Union has formally blocked Atlas deployment on Hyundai and Kia factory floors until a labour-management agreement is reached. There is no current agreement. Every announced timeline depends on resolving a labour dispute in one of the world's most combative automotive union environments — a risk the investor presentations did not prominently address.

DOE selected five companies, including Oklo, for advanced negotiations to receive up to 20 tonnes of surplus weapons-grade plutonium for conversion into advanced reactor fuel — the first time the US government has proposed directing Cold War weapons material into commercial nuclear energy.

The programme converts a long-standing national security liability — plutonium that costs billions to store safely and cannot simply be disposed of — into a strategic asset for next-generation reactor development, with DOE retaining oversight of fuel custody and reactor deployment.

Oklo's Aurora Powerhouse is a compact fast reactor that operates on recycled nuclear material, making it architecturally suited for plutonium-derived mixed oxide fuel — the company was not competing for a generic fuel contract but for fuel that matches its existing reactor design.

Fuel supply uncertainty is the single biggest barrier to advanced reactor commercialisation. Utilities will not sign power purchase agreements for reactors that cannot guarantee fuel access at predictable cost. A government-allocated fuel supply removes that constraint entirely for Oklo's first commercial builds — allowing it to de-risk the commercial timeline in a way no private fuel procurement arrangement could.

The programme transforms the political economy of weapons plutonium: decades of disposal costs become subsidised fuel inputs for next-generation reactors. It also advances non-proliferation objectives by converting weapons material into civilian energy — a strategic framing that gives the programme bipartisan support.

This represents the most direct defence-to-commercial nuclear handoff in US history — Cold War weapons material becoming the input fuel for the commercial reactors that will power the next phase of AI infrastructure.

Oklo (NYSE: OKLO) is publicly listed and the most directly investable beneficiary — the programme selection de-risks its Aurora deployment timeline and reduces the capital required to bring its first commercial unit online. The stock rose 12% in pre-market trading following the announcement.

Second-order opportunities include hot cell and fuel fabrication facility operators (handling weapons-grade material requires specialised infrastructure), newcleo (private, European-listed, potential co-investment vehicle), and the nuclear materials transport and custody firms supporting DOE's material transfer process.

Risk caveat: this is a selection for advanced negotiations, not a final contract. The programme must clear a NEPA environmental review, Congressional notification requirements, and international safeguards obligations under the NPT — each a timeline risk for Oklo's deployment schedule.

BioCirc signed a 7-year agreement to deliver 650,000 CRUs (one tonne CO₂ each) to Microsoft, at 100,000 per year from H2 2026 through 2032, making it BioCirc's largest agreement to date.

Credits come from biogenic CO₂ captured at five Danish biogas plants (Favrskov, Vesthimmerland, Haderslev, Grønhøj, Vinkel), permanently stored via Ineos' Project Greensand CO₂ storage facility in the North Sea — completing a full BECCS value chain entirely within Denmark.

The deal arrives shortly after reports of Microsoft adjusting its CDR procurement; Microsoft subsequently clarified it was adjusting pace and volume, not exiting the market — the BioCirc agreement is the first public contract to confirm that the adjustment was tactical, not structural.

Microsoft is the single largest contracted buyer of durable carbon removals globally, having contracted approximately 45 million tonnes to date. Its purchasing decisions function like central bank signals for the CDR market — setting implicit price floors and determining whether project developers can access infrastructure financing at viable rates. Its re-entry restores the demand signal the market lost during the reported pause.

The BECCS architecture is significant: CO₂ is captured from biomass combustion (which is already part of biogas operations) and stored geologically — combining energy production with permanent removal. This is different from DAC (which consumes energy to remove CO₂) and different from the afforestation CDR Octopus Energy funded three weeks ago. Microsoft choosing BECCS suggests it is diversifying removal pathways, not rotating between them.

BioCirc's CEO noted the system could eventually be replicated across Europe's 1,500+ biomethane production sites— making this deal a template, not a one-off. If it does, BECCS in Europe moves from a niche instrument to a scale-relevant CDR pathway, with significant implications for CO₂ transport and storage infrastructure investment.

Microsoft's re-entry directly de-risks BioCirc's expansion financing and validates the BECCS project finance case for comparable Danish and Nordic operators. These projects have the biomass supply, CO₂ storage geology, and regulatory framework already in place — what they needed was an anchor buyer. Microsoft is that buyer.

Second-order opportunities include North Sea CO₂ transport and storage infrastructure (Project Greensand, Northern Lights, and the emerging European CO₂ shipping network), biomethane logistics (agricultural residue and forestry waste supply chains feeding BECCS facilities), and MRV platform providers that certify BECCS credits to the standard major buyers require.

Risk caveat: BECCS depends on certified sustainable biomass supply — if biomass demand rises faster than verified sustainable supply, the carbon balance of BECCS projects can deteriorate and credit validity can be challenged. The durability of the Microsoft deal depends on BioCirc maintaining supply chain integrity across 7 years of delivery.