Startups

The private companies powering the AI stack

Name: AI startups across the stack
Creator: QAI

The iconic private and venture-backed names mapped across the AI stack — model labs, AI-native apps and agents, custom silicon, robotics and compute infrastructure — each carrying the same web-grounded dossier, last round and valuation trail as every public name. Analyst work product, not advice.

87 startups

Read the thesis before it’s consensus — every board, dossier & note the moment the desk ships it.

Researched & built with QAI proprietary data · QAI Research Framework · QAI Financial Harness · Claude Fable 5 and Opus 5 modelsHow it's built →

87 companies

1X Technologies

$10BTBA

Norway-founded, now Silicon Valley-headquartered (Palo Alto, relocated summer 2025), OpenAI- and EQT-backed humanoid-robotics company selling NEO, a $20k / $499-per-month consumer home humanoid that today runs largely on remote human teleoperation ('Expert Mode') while it gathers data toward autonomy.

IPO: No announced IPO timeline; 1X is focused on NEO commercialization (2026 deliveries) and closing its in-progress Series C, and remains privately, venture-funded

Agility Robotics

Maker of Digit, a bipedal humanoid robot for warehouse/factory logistics, sold mostly as a monthly Robots-as-a-Service subscription; going public via a $2.5B SPAC merger to become the first U.S. pure-play humanoid on public markets.

IPO: H2 2026 — targeted to close before end-2026, pending CCXI shareholder vote + SEC Form S-4 review

AI21 Labs

Israeli enterprise-AI foundation-model lab (Jamba hybrid Transformer-Mamba LLMs + Maestro agent-orchestration platform), private since 2017; last priced at ~$1.4B (2023 Series C, backed by Google + Nvidia), with a widely reported ~$300M Series D (2025) that was never closed, now pivoting hard from selling standalone models to agents after a 61% mid-2026 workforce cut.

IPO: No near-term IPO; a strategic sale / acqui-hire is the more probable liquidity path (2026-2027). Repeated M&A interest (Nvidia, Nebius) rather than IPO-track signals precede any public listing.

Aligned Data Centers

Aligned Data Centers is a private US hyperscale/colocation data-center developer-operator (~5GW operational + planned across ~50 campuses) known for adaptive Delta3 air cooling and DeltaFlow~ liquid cooling, being taken private at a record ~$40B enterprise value by a BlackRock/GIP-Microsoft-MGX-Nvidia consortium.

IPO: None contemplated — Aligned is moving the other direction: it agreed in Oct 2025 to a ~$40B take-private by the AIP/MGX/BlackRock-GIP consortium (expected close H1 2026). No IPO or public listing is planned; the platform stays private under sponsor ownership.

Ambience Healthcare

Ambience Healthcare is a private, VC-backed ambient-AI platform for health systems that auto-generates clinical documentation, coding, and CDI from the patient encounter, expanding from AI scribe into the clinical revenue-integrity stack.

IPO: No public IPO signal; earliest plausible 2028+ given the Jul-2025 Series C and pre-profitability - the fresh $243M provides multi-year private runway

Apptronik

Austin-based humanoid-robot maker whose Apollo general-purpose robot (currently the Apollo 2 prototype generation) is piloting in Mercedes-Benz plants and GXO warehouses, backed by Google DeepMind's Gemini Robotics AI and ~$935M of Series A capital at a ~$5B+ valuation.

IPO: No announced IPO timeline; a listing is unlikely before the Apollo 3 commercial ramp (2027) proves out fleet economics - realistically 2028+ if execution holds. Just raised private capital (Feb 2026), so near-term financing need is covered.

Ayar Labs

Ayar Labs is a private silicon-photonics company building in-package optical I/O (the TeraPHY chiplet + SuperNova external laser) to replace copper interconnect inside AI accelerators; last priced at ~$3.75B in a Mar-2026 Series E and targeting high-volume co-packaged-optics production.

IPO: 2027-2028 (~1-2 years post Series E)

Baseten

Baseten is a fast-growing private AI inference-infrastructure platform that runs open and custom models in production across a multi-cloud GPU fleet, riding the 2025-26 shift of AI spend from training to inference.

IPO: 2027+ at earliest; no S-1, confidential filing, or public IPO signal. Fresh $1.5B Series F (Jun 2026) removes any near-term financing need to go public.

Basis

Basis is a New York AI-agent startup automating end-to-end accounting, tax, and audit workflows for CPA firms; $1.15B private valuation after a $100M Series B in Feb 2026.

IPO: No near-term plan disclosed; earliest plausible 2028+ and only if ARR scales and net retention is proven — freshly capitalized after the Feb 2026 Series B

ByteDance

ByteDance is the privately held Beijing parent of TikTok, Douyin, Toutiao, CapCut, Lark and the Doubao AI stack — the world's most valuable private company, monetizing ~4B+ users chiefly through recommendation-driven advertising and short-video commerce.

IPO: No confirmed filing; recurring Hong Kong / U.S. IPO speculation (2026-2027) that ByteDance has repeatedly declined to confirm. TikTok U.S. JV resolution (Jan 2026) revived listing chatter but no S-1.

Canva

Canva is the world's largest visual-communication / design SaaS platform — a freemium, product-led juggernaut at ~$3.5B revenue (~$4B ARR) and 265M+ monthly users, profitable for eight straight years, now staging a US redomicile and a widely-expected 2026 IPO.

IPO: 2026 (widely expected but unconfirmed; COO Cliff Obrecht: will list 'in the next couple of years'). No exchange or date confirmed.

Character.AI

Character.AI is a private consumer AI-companion/entertainment platform (personalized chatbots, AI video/AvatarFX) that was effectively hollowed out by Google's Aug-2024 $2.7B non-exclusive technology-license-plus-acqui-hire, and is now a restructured, model-agnostic entertainment app fighting user decline and youth-safety litigation.

IPO: No near-term IPO signaled; most plausible endgame is a further strategic sale or continued independent operation on the Google cash cushion, not a public listing

Cognition

Cognition is the maker of Devin, an autonomous AI software engineer, which after acquiring the Windsurf IDE has scaled to a ~$492M ARR run-rate and a ~$26B private valuation as one of the two or three category leaders in agentic AI coding.

IPO: No filed timeline; earliest plausible 2028+ given the May 2026 Series E (~$1B+ raised) and ample private runway - company is scaling ARR, not signaling a listing

Cornelis Networks

Private US high-performance interconnect company building the Omni-Path fabric (CN5000 400G, shipping; CN6000 800G SuperNIC, unveiled Nov 2025) as a non-NVIDIA scale-out network for AI and HPC clusters, backed by Intel Capital and funded by DOE-NNSA and IARPA.

IPO: No public IPO timeline or filing signals as of 2026-07; given sub-scale capitalization (~$85-139M lifetime raised) and reliance on small rounds, a strategic acquisition (AMD, Intel, a server OEM, or a networking vendor) is a more probable liquidity path than an IPO.

Cresta

Cresta is a private Sunnyvale-based generative-AI contact-center platform (real-time agent assist, conversation intelligence, QM, and virtual agents) co-founded by Sebastian Thrun, last funded at a $125M Series D (Nov 2024).

IPO: No public IPO timeline; not expected before 2026+ given the undisclosed/flat 2024 mark and continued private-market financing posture

Anysphere (Cursor)

Anysphere builds Cursor, the AI-native code editor/agent that became one of the fastest-scaling software companies ever (~$4B annualized revenue in under four years), now the subject of an announced $60B all-stock SpaceX acquisition after a $29.3B Series D.

Acquired by SpaceX (2026): ~$60B all-stock (announced Jun 16 2026, closing ~Q3 2026)

Acquired · SpaceXAppsAI

Context31

CXMT

ChangXin Memory Technologies

$350BH2 2026

China's only large-scale DRAM maker and the world's #4 memory supplier, riding the 2026 AI-memory supercycle to explosive registration-disclosed earnings as it prepares a record ~$4.3B Shanghai STAR Market IPO at an analyst-floated CNY 2-3 trillion valuation.

IPO: Late-Aug to Sept 2026 STAR Market debut (~3-4 months post-approval); listing-committee review cleared and registration filed 2026-05-27, CSRC registration approved June 2026

Decagon

Decagon builds AI 'concierge' customer-service agents (chat, email, voice) for large enterprises, differentiated by its Agent Operating Procedures (AOP) framework; a ~3-year-old startup that tripled to a $4.5B valuation in under six months.

IPO: Not before 2028; no S-1 or confidential filing disclosed. The Mar 2026 employee tender (early secondary liquidity at $4.5B) typically signals a long private runway rather than an imminent listing.

DeepSeek

China's breakout open-weight frontier-model lab, spun out of quant fund High-Flyer, that matches Western frontier models on agentic/reasoning benchmarks at ~10x lower API cost -- privately held, first outside capital raised in 2026 at up to a ~$59B valuation.

IPO: No public timeline -- mainland-China private company, research-first mandate; founder has repeatedly deprioritized commercialization over an IPO. Any listing would most plausibly be a domestic (STAR Market / HK) venue given US-China decoupling, not a US ADR.

d-Matrix

Santa Clara AI-inference chip startup building digital in-memory compute (DIMC) accelerators (Corsair) to undercut Nvidia on inference cost, latency, and energy; Microsoft- and Temasek-backed, ~$2B private valuation.

IPO: Not before 2027; IPO optionality flagged after the oversubscribed Series C, but no S-1 and CEO states no intent to sell

ElevenLabs

The category-leading AI voice/audio platform (TTS, dubbing, and voice agents) that ~95% of first-time voice-AI buyers start with, scaling ARR from ~$100M (2024) to $500M+ (2026) and last priced in a primary round at $11B (Feb 2026), with secondary talks circling ~$22B.

IPO: No confirmed date or venue; CEO Mati Staniszewski has said the company is building toward an IPO but named no timeline. Crossover/PE entrants (BlackRock, Wellington, D.E. Shaw, Schroders) in the Series D third close and ~$22B secondary talks point to a listing readiness likely 2027+, not near-term. Reports of a specific date or a Warsaw Stock Exchange listing are unconfirmed.

EvenUp

EvenUp is a private San Francisco AI startup (Series E, ~$2B+, Oct 2025) building the leading AI platform for U.S. plaintiff-side personal-injury law — auto-drafting demand letters, medical chronologies and case analyses, now extending into a Pre-Litigation-as-a-Service (PLAAS) model.

IPO: No public timeline. Fresh off a $150M Series E (Oct 2025) and pushing into the PLAAS services expansion; a public listing is most plausibly 2+ years out (2028+), and a strategic acquisition by a legal-data incumbent (RELX/Thomson Reuters) is a live alternative exit.

EV Group

EV Group is a privately held, family-owned Austrian semiconductor-equipment maker that quietly monopolizes wafer bonding (~82% share) and is a front-line enabler of hybrid bonding and nanoimprint lithography for 3D-IC and advanced packaging.

IPO: None disclosed — family-owned (E. Thallner family) and self-funded since its 1980 founding; no IPO, S-1, or exit process signaled.

Fireworks AI

Fireworks AI is a private, fast-growing GenAI inference-and-fine-tuning platform (founded 2022 by ex-Meta PyTorch leads, incl. CEO Lin Qiao) that serves open-weight models at high speed and low cost, riding the inference-token boom to a reported ~$800M annualized revenue run-rate in mid-2026.

IPO: No public IPO timeline; company is focused on private growth financing (reportedly raising a ~$15B private round as of mid-2026, not yet closed). A public listing is not expected before the ~$15B round matures and revenue scales past a $1B+ run-rate — realistically 2027+ at the earliest, contingent on AI-infra market conditions.

Fivetran

Fivetran is the leading automated data-movement (ELT/CDC) platform; on June 1, 2026 it completed an all-stock merger with dbt Labs to form a private data-infrastructure company (~$600M combined ARR) positioning as the governed 'data layer for AI agents.'

IPO: No filed timeline; an eventual IPO of the combined ~$600M-ARR Fivetran+dbt entity is contemplated when public markets reopen (2027+ at the earliest). No S-1, no confidential filing disclosed.

FluidStack

London-founded (relocating global HQ to NYC) private 'neocloud' that rents Nvidia GPU compute at scale via an owned-datacenter private-cloud business plus an asset-light marketplace; catapulted from a niche broker to AI infrastructure heavyweight by a Nov-2025 $50B build-out partnership with Anthropic, and in talks (Apr-2026) to raise $1B at an ~$18B valuation.

IPO: None disclosed; focused on a private growth round (~$1B at ~$18B in talks, Apr-2026). A future IPO is plausible on neocloud comps (CoreWeave, Nebius listed) if the AI-capex cycle holds, but no S-1 or timeline exists.

G42 (Group 42) is Abu Dhabi's state-aligned AI national champion: a private holding company controlled by the UAE's National Security Adviser, spanning AI compute/cloud, data-center infrastructure, applied-AI analytics, space/geospatial and healthtech, and a primary UAE recipient of US advanced-chip export allocations.

IPO: Rumored

Glean

Glean is a privately held enterprise 'Work AI' platform (Palo Alto, founded ~2019 by ex-Google distinguished engineer and Rubrik co-founder Arvind Jain) that indexes a company's apps into a permissions-aware knowledge graph to power cross-app search, an AI assistant, and no-code AI agents; last priced at a $7.2B valuation (Series F, Jun 2025) and reporting ~$300M ARR as of May 2026 (+89% YoY).

IPO: 2027+ (no filing; CEO Arvind Jain has framed an IPO as a long-term possibility, not a near-term event; Glean remains venture-funded post-Series F with ample runway)

Graphcore

UK AI-accelerator pioneer (the IPU) that flamed out as an independent NVIDIA challenger, was bought cheaply by SoftBank in 2024, and is now a captive silicon unit inside SoftBank's Arm+Ampere+Graphcore vertical AI-compute stack aimed at Stargate.

IPO: None - wholly-owned SoftBank subsidiary since Jul 2024; delisted from the merchant market as an independent entity, no IPO path while captive to SoftBank/Stargate

Harvey

Harvey is a domain-specific generative-AI platform for elite law firms and enterprise legal/professional-services teams, building agentic workflows for research, drafting, contract analysis, due diligence and litigation on top of frontier foundation models (OpenAI, Anthropic, Google) it routes across and tunes for legal.

IPO: No S-1 and no confirmed IPO plan; earliest plausible listing 2027-2028 if hyper-growth persists and gross-margin/profitability visibility matures. Still raising large private rounds and executing employee secondaries in lieu of a public offering.

Pre-IPOAppsAI

Context14

Loading more…

Analyst work product, NFA — always verify. Prices are real (delayed) trends for context.