The first humanoid company is going to Wall Street. Oil is powering AI. Nuclear is back.

Agility signed a definitive SPAC merger agreement on June 24 valuing the company at $2.5 billion, expected to raise $620M+ in gross proceeds through a combination of Churchill's trust account and a Foxconn-led PIPE — the largest capital raise in humanoid robotics history.

Upon closing, Agility will trade on Nasdaq as AGLT, becoming the first US-listed pure-play humanoid robotics company with active commercial deployments and more than $300 million in forward orders for its Digit v5 platform.

Backers include NVIDIA, Amazon, SoftBank Vision Fund 2, Foxconn, and DCVC; NVIDIA's dual role as investor and technical partner — providing AI compute and simulation tools for robot training — signals that Agility's platform is becoming a reference point for the physical AI stack.

The listing creates the first public benchmark valuation for commercial humanoid robotics — a market Goldman Sachs projects could reach $38 billion by 2035. AGLT's debut price will immediately reprice private competitors including Figure AI, 1X Technologies, and NEURA Robotics.

NVIDIA's participation and Foxconn's manufacturing stake together point toward a structural platform dynamic: AI compute on one side, contract manufacturing scale on the other, with Agility's robot in the middle — a position no other humanoid company currently holds simultaneously.

The SPAC comes as humanoid supply is beginning to outpace demonstrable commercial demand — Agility's public market debut is a deliberate signal that capital market discipline, not just venture backing, will be the mechanism that separates real deployment companies from demonstration companies.

Churchill Capital Corp XI (NASDAQ: CCXI) is the current listed vehicle; investors can access the trade through CCXI before AGLT begins trading, subject to redemption and SPAC closing risks. Foxconn's manufacturing partnership capability is the supply chain angle to watch post-listing.

Agility has not disclosed revenue figures; the $2.5B valuation is predicated on future deployment scale and the data flywheel from live commercial operations, not current cash flows. The $300M+ forward order book provides some revenue visibility, but commercialisation timelines remain the key risk.

Risk caveat: SPAC structures have historically underperformed traditional IPOs post-listing, and the deal remains subject to redemption risk. The gap between commercial partnership announcements and recurring humanoid revenue is where the post-listing story will be written.

Chevron and Microsoft signed a 20-year power purchase agreement on June 22 for Project Kilby — a dedicated 2.67GW natural gas facility on 2,000+ acres in Reeves County, West Texas, structured to supply Microsoft's AI data centre operations off-grid, bypassing ERCOT entirely.

The off-grid architecture is a direct response to the grid connection crisis: the US interconnection queue now holds more than 2,700 GW of pending requests, with multi-year approval timelines that threaten the AI data centre buildout. Kilby sidesteps the queue and delivers power directly to Microsoft.

Kilby draws on Permian Basin gas from Chevron's existing production, where volumes routinely exceed pipeline takeaway capacity — giving the project a structural cost advantage through discounted stranded gas, with FID targeted by end of 2026 and a mid-teen return target.

Project Kilby is one of the first examples of a major oil company becoming a direct AI infrastructure provider — not selling fuel to a utility, but building and owning a power plant contracted directly to a hyperscaler. This is a structural shift in the energy-tech value chain with implications for every integrated energy company.

The 20-year PPA and mid-teen return target position this as infrastructure-grade investment, not a commodity energy play — validating that AI power demand is being underwritten at the same credit quality as regulated utility assets. This changes how energy companies should model AI demand in their capital allocation frameworks.

AI data centre power demand is projected to double US data centre electricity consumption to 77 GW by 2030 per BloombergNEF, and hyperscalers including Google, Amazon, and Microsoft have collectively committed more than $500 billion in capex. Kilby is the clearest signal yet that dedicated co-located power is the structural solution to the grid bottleneck.

The primary listed exposures are Chevron (NYSE: CVX) as developer and owner of Kilby and GE Vernova (NYSE: GEV) as primary equipment provider. Engine No. 1 holds an option to take 50% of Kilby's ownership — its participation is not yet confirmed.

Kilby's FID by end of 2026 creates a near-term catalyst for Chevron; if committed, it represents a multi-billion-dollar capital deployment in a segment that is explicitly decoupled from oil and gas price cycles — the structural diversification story Chevron has been signalling to investors.

Risk caveat: Kilby is not yet at FID and remains subject to financing conditions, regulatory approvals, and environmental reviews. Natural gas pricing and Permian Basin gas availability will materially affect 20-year economics, and political exposure around fossil fuel-powered AI infrastructure is increasing.

DOE's Office of Energy Dominance Financing issued a conditional $17.5 billion loan commitment on June 23 to finance long-lead components for up to 10 Westinghouse AP1000 reactors across five projects — the largest US nuclear financing commitment in a generation, designed to restart domestic supply chain manufacturing.

Westinghouse has signed LOIs with seven utilities with identified project sites. Policy advisory firm Capstone identifies the most likely candidates as Dominion Energy, DTE Energy, WEC Energy Group, PSEG, and Entergy, which have existing NRC licensing work and suitable sites. DOE expects to name recipients in the second half of 2026.

Energy Secretary Chris Wright confirmed that the loans have drawn direct interest from AI data centre hyperscalers and energy companies — confirming that this nuclear investment is as much an AI infrastructure play as a decarbonisation programme.

This is the structural intervention the nuclear supply chain has needed for three decades. Long-lead components currently take 5–7 years to procure, the single biggest barrier to new nuclear build timelines. The bulk purchase structure — locking in component prices before individual projects proceed to FID — changes the incentive structure for component manufacturers to rebuild domestic capacity.

The AP1000's modular factory construction approach — the same methodology used in Vogtle Units 3 and 4, the only large nuclear reactors built in the US in 30 years — is the template this programme replicates at scale. Vogtle is now operating; the DOE is betting it can recreate those units faster and cheaper at volume.

The programme is directly linked to the Trump administration's EO 14302 mandate to have 10 large reactors under construction by 2030 — creating a policy-backed procurement cycle that gives utilities a credible reason to commit to nuclear at a time of unprecedented electricity demand.

The primary listed supply chain plays are Cameco (NYSE: CCJ), which owns 49% of Westinghouse alongside Brookfield Asset Management (NYSE: BAM), and GE Vernova (NYSE: GEV). Long-lead reactor component manufacturers including BWX Technologies are also positioned as direct beneficiaries.

The programme creates a visible, policy-backed order book for nuclear components — the type of de-risked revenue pipeline that historically attracts infrastructure capital. Utilities selected as loan recipients in H2 2026 will be the next catalyst.

Risk caveat: The commitment is conditional: utilities must satisfy technical, legal, environmental, and financial requirements before funds are disbursed. Cost overruns at Vogtle — the most recent US AP1000 build — were significant, and replicating that project faster does not eliminate execution risk.

Trump signed two quantum executive orders on June 22 — one directing a national quantum innovation push including a 2028 fault-tolerant computing target; the second mandating post-quantum cryptography implementationacross all federal high-value systems by end of 2030.

DOE announced Quantum Genesis on June 23: a funded, multi-agency initiative targeting the world's first fault-tolerant quantum computer for scientific research by 2028, centred on the DOE Q Competition, a National Quantum Supercomputing User Facility, and targeted quantum application R&D.

The combined EO and Quantum Genesis package positions quantum as the third pillar of US national computational infrastructure alongside AI and HPC — the first time the federal government has set a specific technical deliverable with a funded institutional mechanism behind it.

The 2028 fault-tolerant target is credible in a way commercial roadmaps are not: it is a technically specific deadline from DOE's National Labs, which have the cryogenic, photonic, and ion-trap hardware infrastructure required to pursue it — building on more than $1 billion invested in National Quantum Information Science Research Centers since 2020.

The post-quantum cryptography mandate creates a hard procurement cycle: every federal agency must identify a PQC migration lead by late July 2026, and complete implementation of quantum-resistant standards by end of 2030–2031. This is a direct near-term revenue signal for PQC vendors, irrespective of whether fault-tolerant hardware arrives by 2028.

The executive orders explicitly frame this as a US-China competition — following China's own quantum commitments in the 14th Five-Year Plan. Federal quantum capital is now formally racing Chinese state quantum capital in a strategic technology, the same pattern that accelerated the US-China semiconductor contest.

The Q Competition will require hardware from private vendors. The leading listed candidates for DOE partnerships include IonQ (NYSE: IONQ) and Rigetti Computing (NASDAQ: RGTI). Private competitors Alice & Bob and QuEra Computing (backed by Google and DARPA) are also well-positioned but not yet investable via public markets.

The post-quantum cryptography mandate creates a 2026–2030 procurement cycle for cybersecurity companies with PQC capabilities — a nearer-term revenue event than the 2028 hardware target and accessible through existing cybersecurity holdings.

Risk caveat: No fault-tolerant quantum computer exists today, and independent experts generally place practical fault tolerance at 2028–2033 at the earliest. The DOE Q Competition is a challenge prize, not a guaranteed delivery — the 2028 deadline is a policy target, not a product commitment.

Hyundai Motor Group is acquiring SoftBank's remaining 9.65% stake in Boston Dynamics for $325 million, following SoftBank's exercise of a contractual put option — converting Boston Dynamics into a wholly owned Hyundai subsidiary pending final board approvals across Hyundai Motor, Kia, Hyundai Mobis, and Hyundai Glovis.

The deal price implies a Boston Dynamics valuation of ~$3.37 billion at the contractual floor — a fraction of the $19–22 billion implied by Hyundai Glovis's internal August 2025 investment valuation, which reflects the surge in physical AI market premiums since the original 2021 transaction.

Atlas humanoid deployments at Hyundai facilities are beginning in 2026, with expansion to the Georgia Metaplant planned for a later phase; Google DeepMind is separately working with Boston Dynamics to integrate advanced AI reasoning models into Atlas, providing a software stack alongside Hyundai's hardware and manufacturing access.

Full ownership removes the put option constraint that limited Hyundai's ability to drive Boston Dynamics' roadmap. Under sole ownership, Hyundai can directly align Atlas development and production with its automotive manufacturing operations — the most commercially realistic near-term deployment environment for industrial humanoids anywhere.

Boston Dynamics' primary advantage over Agility, Figure AI, and NEURA is access to a vertically integrated customer in Hyundai's own factories, with known layouts, high-volume production cycles, and a supply chain link through Hyundai Mobis for actuators and components. No other humanoid company has this structural shortcut to commercial deployment.

SoftBank's exit is a deliberate capital reallocation: with a reported $41 billion position in OpenAI and a new AI infrastructure venture called Roze AI, SoftBank is exiting physical robotics to concentrate capital in AI software and data centre infrastructure — a signal about where Japanese sovereign-adjacent capital is placing its next decade.

The primary listed exposure is Hyundai Motor (KRX: 005380), which is consolidating one of the most recognisable robotics brands globally at a transaction price that likely understates Boston Dynamics' current value by a factor of five or more relative to physical AI sector premiums.

The $3.37B implied valuation at the put option price versus the $19–22B internal estimate creates a potential IPO arbitrage: if Boston Dynamics lists in 2027–2028 — as the original 2021 deal anticipated — the valuation gap could represent significant upside currently captured entirely by Hyundai.

Risk caveat: Boston Dynamics has not yet generated meaningful commercial revenue from Atlas at scale. CEO Robert Playter has stated Atlas must learn new tasks within a day or two and reach 99.9% task reliability before broad factory deployment — a bar not yet publicly cleared. The deal also has not been formally confirmed.