Know the franchise before the roadshow

TAE Technologies, Inc.~$6B

Beam-driven field-reversed-configuration (FRC) fusion developer targeting aneutronic hydrogen-boron (p-B11) fuel; ~$1.3B+ raised over ~25 years, now going public via a ~$6B all-stock reverse merger with Trump Media (DJT), expected to close mid-2026.

General Fusion (General Fusion Group Ltd. — pending Nasdaq: GFUZ)~$1B

Canadian magnetized-target-fusion (MTF) developer whose LM26 demo machine mechanically compresses a lithium-lined plasma — no superconducting magnets or lasers — pursuing a Nasdaq listing via a ~$1B SPAC after a near-death 2025 cash crunch; commercial net-energy is a mid-2030s aspiration, so this is a speculative, financing-gated, non-US (analysis-only) name.

Finland-based superconducting full-stack quantum-computer maker — Europe's hardware leader, selling on-premises systems to HPC/supercomputing centers — closing a ~$1.8B de-SPAC (Nasdaq: IQMX) to become the first listed European quantum pure-play.

Anthropic~$965B

Frontier AI safety lab behind the Claude model family and Claude Code; #1 most-valuable private AI startup, confidentially filed for a ~Oct-2026 IPO.

Lambda~$5.9B

Lambda is a developer-first AI "neocloud" that rents NVIDIA GPU capacity (on-demand and reserved 1-Click Clusters) for model training and inference, now anchored by multibillion-dollar Microsoft and NVIDIA contracts.

Pasqal SAS (pending Nasdaq: BBCQ)~$2B

French neutral-atom quantum computing pure-play — co-founded by 2022 Nobel laureate Alain Aspect, backed by Temasek/Orange/BPIfrance — going public via a $2.0B SPAC merger with Bleichroeder Acquisition Corp II (Nasdaq: BBCQ), expected to close H2 2026.

Crusoe Energy~$23.6B

Crusoe is a vertically integrated AI-infrastructure company that builds and operates gigawatt-scale data centers (including OpenAI's flagship Stargate campus in Abilene, Texas) by pairing its own power generation with GPU cloud, after pivoting fully out of flared-gas Bitcoin mining in 2025.

Databricks~$134B

Databricks is the leading data + AI 'lakehouse' platform — unifying data engineering, warehousing, ML and generative-AI agents on open formats (Delta/Iceberg, Unity Catalog, Mosaic AI) across AWS, Azure and GCP.

Anduril Industries~$61B

A software-defined defense prime that builds autonomous drones, counter-UAS, subs and combat aircraft and ties them together with its Lattice AI command-and-control OS, selling finished products at fixed price rather than cost-plus programs.

Cohere~$20B

Toronto-based enterprise LLM platform — Command generative models, Embed/Rerank retrieval, and the North agent platform — sold to regulated enterprises and governments on a security-, multilingual-, and data-sovereignty-first positioning rather than consumer scale.

QuEra Computing~$1B

Boston neutral-atom quantum-computing pure-play, commercializing the Lukin (Harvard) / Vuletić (MIT) Rydberg-atom array — analog (Aquila on AWS Braket) today, digital error-corrected (96 logical qubits demonstrated) en route to its fault-tolerant Libra system on Braket by 2028; Google-, SoftBank-, and NVIDIA-backed.

Shield AI builds Hivemind, an AI pilot that flies aircraft and drones autonomously in GPS- and comms-denied combat environments, plus its own V-BAT surveillance drone and the in-development X-BAT VTOL autonomous fighter jet.

Atom Computing, Inc.Undisclosed

Neutral-atom (optical-tweezer) quantum-computing startup building full-stack fault-tolerant machines — held the 1,180-qubit Phoenix record, demonstrated neutral-atom error correction, and is co-delivering the 50-logical-qubit Magne system with Microsoft to Denmark's QuNorth.

OpenAI~$852B

OpenAI is the frontier AI lab behind ChatGPT and the GPT/o-series models, monetizing via consumer subscriptions, a developer API, and enterprise deals while spending tens of billions on compute.

Figure AI builds general-purpose humanoid robots (Figure 03) powered by its in-house Helix vision-language-action model, deployed today on commercial factory floors (BMW Spartanburg) and ramping toward home use via its BotQ manufacturing facility.

Together AI~$8.3B

Together AI is an "AI-native cloud" that runs, fine-tunes, and trains 200+ open-source models on NVIDIA GPUs — pairing a serverless OpenAI-compatible inference API with dedicated GPU-cluster rental, built by the research team behind FlashAttention and RedPajama.

Best-funded private fusion company globally (MIT spinout) building a high-temperature-superconducting (HTS) tokamak — SPARC demo targeting net energy gain (Q>1) ~2027, then the ARC ~400 MWe power plant in Chesterfield County, Virginia for grid power in the early 2030s.

Perplexity AI~$20B

Perplexity is an AI-native answer engine that returns cited, synthesized answers instead of blue links, now extending into an agentic browser (Comet) and autonomous task agent (Computer) to own the surface where an AI agent starts — and increasingly finishes — a task.

PsiQuantum Corp.~$7B

Photonic, fault-tolerant quantum-computing pure-play building utility-scale (~1M physical-qubit) machines on silicon-photonic chips fabricated at GlobalFoundries, anchored by US (Illinois) and Australian (Queensland) government-backed fab/deployment sites; ~$7B last-round valuation (Series E, Sep 2025).

SandboxAQ~$5.75B

Alphabet-spinout enterprise AI/quantum company selling "Large Quantitative Models" (physics-grounded AI) plus post-quantum cryptography (AQtive Guard) and quantum-sensing navigation (AQNav); last marked ~$5.75B (Series E, Apr-2025), with a June-2026 $500M U.S. CHIPS materials-discovery award that gives the government a minority equity stake.

Blue Origin~$130B

Jeff Bezos's privately held space company building the reusable New Glenn heavy-lift rocket, BE-4 engines, the New Shepard suborbital vehicle, NASA Blue Moon lunar landers, and proposed orbital-data-center constellations — the principal Western challenger to SpaceX.

Groq~$6.9B

Groq designs the LPU, a deterministic SRAM-based AI-inference accelerator, and sells low-latency, low-cost token generation through GroqCloud and on-prem GroqRack — and, since December 2025, licenses that inference IP to Nvidia.

Mistral AI~€11.7B

Paris-based frontier AI lab building open-weight and enterprise LLMs (Mistral Large 3, Medium 3.5, Ministral 3) plus the Vibe assistant/coding agent, positioned as Europe's sovereign alternative to OpenAI and Anthropic.

SambaNova Systems~$11B

A Palo Alto AI-chip company whose Reconfigurable Dataflow Unit (RDU) and full-stack SambaCloud sell enterprise/sovereign LLM inference as an alternative to NVIDIA GPUs, now repositioned around fast, memory-efficient agentic inference.

Helion Energy, Inc.~$15.5B

Sam Altman-chaired private fusion developer using a field-reversed-configuration, pulsed-magnetic-compression machine that captures electricity directly from the plasma — racing a 50MW Microsoft power purchase agreement (Orion plant, target 2028) on the aneutronic deuterium–helium-3 fuel cycle, the hardest fuel cycle in fusion.

UK spherical-tokamak fusion developer with a dual model: a long-dated bet on commercial fusion (ST-E1 pilot, early-2030s) funded today by a genuinely commercializing HTS-magnet business (TE Magnetics) that already has a £70M UK STEP contract and US defense (DARPA/General Atomics) magnet revenue. Non-US issuer — analysis-only.

Ilya Sutskever's research-only AI lab building a safe superintelligence as its first and only product, with no API, no chatbot, and no revenue until the mission is solved.

Etched~$5B

Etched designs Sohu, a transformer-only inference ASIC (TSMC N4P), now sold as a full 'one-stop' datacenter system — Frontier Inference Clusters (chip + racks + cooling + networking + software) with Low-Voltage Inference and Cluster-Scale Memory — betting that hard-wiring the transformer into fixed-function silicon wins inference on tokens-per-dollar and tokens-per-watt.

Reflection AI~$25B

US open-weights frontier AI lab founded by ex-DeepMind researchers, positioned as America's answer to Chinese open-weights models; newest SpaceX Colossus 2 anchor tenant (~$6.3B compute deal) — PRIVATE, not ownable.

Starcloud~$1.1B

Orbital AI-datacenter pure-play that flew the first data-center-class GPU (an Nvidia H100) in low-Earth orbit; PRIVATE, pre-revenue, fastest YC company to reach unicorn status — a venture-watch anchor for the space-compute theme, not an ownable long.