Private Markets Update & Deal Flow

Anduril Industries announced it raised USD 5B in a Series H round led by Thrive Capital and Andreessen Horowitz, pushing its valuation to USD 61B — nearly double its previous USD 30.5B mark. CEO Brian Schimpf confirmed revenues of USD 2.2B in 2025, also roughly double the prior year. The company will invest aggressively in manufacturing, R&D, and infrastructure, including its Arsenal-1 high-rate production facility and the Lattice AI command-and-control platform.

Separately, Anduril and the US Army successfully tested launching the Altius 700 drone from an AH-64E Apache helicopter for the first time during the CFWE26 military exercise. The Altius 700 is a long-range loitering munition capable of travelling hundreds of kilometres — well beyond the reach of the Apache's conventional missiles.

The strategic context is clear. After the Russian invasion of Ukraine exposed helicopter vulnerability near the front line, the US military has shifted to stand-off warfare — keeping rotorcraft at safe distances while drones do the dangerous work. Comparable systems from L3Harris and Israel's Spike NLOS reflect the same doctrine. Key risk: dependence on GPS and comms links that adversaries could jam makes future autonomous navigation upgrades critical.

The signal: Anduril has now raised more than USD 11B in total. The combination of manufacturing scale, Lattice software, live battlefield validation, and a USD 20B army contract makes it the defining pre-IPO defence tech position of this cycle.

Blue Origin is exploring external investment for the first time in its 25-year history as it accelerates rocket launches and seeks to compete more directly with SpaceX. CEO Dave Limp reportedly told employees that achieving the company's ambitions will require a lot of capital — more than could realistically come from a single backer, even with continued support from founder Jeff Bezos.

The discussions follow New Glenn's successful orbital debut in early 2025 and its first booster landing later that year. Blue Origin is now targeting up to 100 launches per year in the long term, competing for commercial satellite contracts and major government projects, including NASA's Artemis lunar programme.

Renewed investor enthusiasm ahead of SpaceX's anticipated IPO — which could value that company above USD 1.75T — has created what Blue Origin sees as a favourable fundraising window. Limp also noted outside rounds could help employees monetise stock options in structures similar to those at SpaceX and OpenAI. Despite the possibility of outside investors, Bezos is not expected to cede control and no formal IPO plans have been announced. The company's longer-term vision extends to its TeraWave satellite communications network.

Worth watching: Blue Origin opening to outside capital after 25 years of sole Bezos funding is a structural shift. The renewed investor appetite ahead of SpaceX's IPO may give it a rare window to raise at attractive terms.

Anthropic became the focal point of a major standoff with the US government after reports that Claude was active inside US military infrastructure during strikes on Iran in early 2026. The model had already been integrated into classified Pentagon systems for intelligence analysis, operational planning, and logistical support before the conflict escalated.

The dispute crystallised when Anthropic refused Pentagon requests to remove certain safeguards. CEO Dario Amodei publicly maintained two red lines: no mass surveillance of American citizens and no fully autonomous weapons systems. The Pentagon argued these restrictions were too limiting and reportedly demanded broader all lawful purposes access.

Tensions escalated when the Trump administration moved to phase out Anthropic's systems and classified the company as a supply-chain risk. In practice, replacing deeply embedded AI systems proved far harder than anticipated — reportedly taking months. The case exposed the structural dependency governments can develop on commercial AI providers, with Claude embedded in military workflows through platforms linked to Palantir Technologies and Project Maven.

The core question: not whether private AI companies support defence — but who retains control once models are embedded in military systems. The answer will shape every AI defence contract written from here on.

SpaceX has signed a major agreement with Anthropic to provide access to its Colossus 1 AI supercomputer cluster in Memphis — more than 220,000 Nvidia GPUs and over 300 megawatts of computing power, previously used to train xAI's Grok models.

The deal is estimated to generate USD 5–6B in annual revenue for SpaceX — significant given that xAI had been burning approximately USD 1B per month expanding its AI infrastructure. Leasing computing capacity to Anthropic transforms a costly operation into a major revenue stream.

The agreement positions SpaceX as a new neocloud provider, competing directly with Microsoft Azure, Amazon AWS, Oracle, and Google Cloud. AI compute is rapidly becoming a distinct third business segment for SpaceX alongside launch services and Starlink.

The trend: SpaceX is quietly becoming a three-business conglomerate — launch, satellite internet, and now AI cloud. The Colossus deal shows how large the margin opportunity in AI compute is relative to even the most capital-intensive rocket programmes.

OpenAI is allowing employees to sell up to USD 30M each in shares through a new private financing round — one of the largest individual employee liquidity events ever seen in the technology industry. This follows the 2023 tender offer, during which employees collectively sold around USD 6.6B worth of shares.

The scale of these payouts reflects the fierce global competition for top AI engineers and researchers. Equity packages at leading AI labs are now comparable to large startup exits or IPO windfalls. OpenAI is increasingly relying on private share sales as the primary mechanism to reward and retain talent without going public.

Repeated liquidity rounds are fuelling speculation about a future IPO, although OpenAI appears to prefer remaining private while continuing to raise capital and offering periodic share sales. The move highlights how dramatically AI valuations have increased, with private investors continuing to pour billions into OpenAI, Anthropic, and Moonshot AI.

The shift: AI lab equity has become the most valuable private compensation in history. Repeated tender rounds at escalating valuations are the new IPO — and private investors, not public market shareholders, are capturing the bulk of the gains.

Cerebras is preparing to significantly increase the size and pricing of its IPO after investor demand reportedly exceeded available shares by more than 20 times. The company is considering raising its price range to USD 150–160 per share, up from the original USD 115–125, while also increasing shares offered from 28M to 30M.

At the top of the revised range, Cerebras could raise approximately USD 4.8B, making it the largest IPO globally so far in 2026. The strong demand reflects continued enthusiasm for AI infrastructure and semiconductors — particularly companies positioned outside Nvidia's dominance.

Cerebras develops specialised AI processors optimised for inference workloads — the computations used when AI models generate responses for users. As the market shifts from training toward large-scale deployment, Cerebras benefits from growing demand for chip architectures better suited to inference tasks. The IPO marks a comeback after the company postponed its 2024 listing due to a national security review tied to G42, which once represented more than 80% of its revenue. That review was cleared, and Amazon and OpenAI are now major customers.

Why it matters: The AI market is shifting from training to inference at scale. Cerebras — with Amazon and OpenAI as anchor customers and a cleared security review — is well positioned to capture that wave as the primary alternative to Nvidia's ecosystem.

The US Chamber of Commerce, citing research by Rhodium Group, has warned that China is accelerating industrial dominance through massive state support, deeper supply-chain control, and rapid expansion into advanced technologies including robotics, AI, biomanufacturing, and nuclear fusion. Beijing has moved from targeted industrial policies to what the report calls an industrial policy of everything.

The report argues that many governments underestimated earlier Chinese strategies — such as Made in China 2025 — and responded too slowly. China is now extending its reach into chemicals, machinery, industrial equipment, and robotics, where Western economies have traditionally held advantages. China's goods trade surplus has roughly doubled to approximately USD 2T since 2019, supported by heavy state backing and weak domestic demand.

Chinese companies are also globalising rapidly, with overseas revenues for the country's top 500 firms reaching levels comparable to large US multinationals. Beijing is increasingly deploying export controls and economic pressure to tighten its strategic grip. Germany, Japan, and South Korea are seen as most exposed given their manufacturing-heavy economies. Without stronger industrial and trade responses, China could dominate many high-tech sectors within a few years.

The signal: China is not just catching up — it is building a parallel industrial system and exporting it globally. More policy-driven and subject to regulatory unpredictability, but the direction of travel and the speed of execution are unmistakably clear.