Skip to content
Reported Confidence: medium A · Pure-play Neocloud

Together AI

AI-native cloud focused on inference and training of open models; runs 'Frontier AI Factories' with partners (5C) and sells serverless/dedicated inference plus GPU clusters.

www.together.ai · Record collected 25 Sept 2026

Profile from NeoCloud Observatory research (snapshot 2026-09-25); every figure links to its original source.

Profile

Headquarters
San Francisco, California, United States
Founded
2022
Ownership
private
Operating regions
North America
Deployment model
inference as a service, dedicated clusters, ai platform software
GPU vendors
NVIDIA
GPU generations
NVIDIA Blackwell, NVIDIA Blackwell Ultra
Networking
Not yet recorded
Storage
Not yet recorded
Cooling
Not yet recorded
Energy strategy
third party — Partner facilities.
Platform / control plane
Serverless and dedicated inference APIs for open models
Business model
Open-model inference is the core product.
Differentiation
Not yet recorded

Sources: TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation, Business Wire (Together AI) — Together AI Raises $800 Million at $8.3 Billion Valuation

Capacity register

Each sourced figure with its maturity and MW type. Figures of different types are not added together.

No sourced capacity figures recorded.

Projects

ProjectLocationStatusCapacity
Together AI / 5C Maryland AI factory

B200 Participants: 5C (partner). Source: NeoCloud Observatory research.

Source: NVIDIA Newsroom — NVIDIA and Partners Build America's AI Infrastructure
United States operational Not sourced
Together AI / 5C Memphis AI factory

GB200/GB300; 'coming online soon' Participants: 5C (partner). Source: NeoCloud Observatory research.

Source: NVIDIA Newsroom — NVIDIA and Partners Build America's AI Infrastructure
Memphis, Tennessee, United States under construction Not sourced

Disclosed customer contracts

No disclosed customer contracts recorded.

Financing & M&A

DateTypeAmountDetail
1 Jul 2026 equity $800.0M Together AI — Series C

Investors: Aramco Ventures, Vista Equity Partners, General Catalyst, Emergence Capital, NVIDIA, Salesforce Ventures, March Capital, Pegatron, Schneider Electric, Lux Capital

Source: TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation
20 Feb 2025 equity — Together AI — Series B $3.3B valuation 17 months before Series C (per TechCrunch); amount not captured. Source: TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation

Neocloud Observatory

Evidence-scored analysis

Category A · Pure-play Neocloud · data quality 59.6/100 · snapshot 25 Sept 2026

Capacity by maturity (MW)

Each stage is reported separately and never added together. Announced GPUs never count as operational.

  • Operational Unknown
  • Under construction Unknown
  • Contracted Unknown
  • Secured Unknown
  • Announced Unknown
  • Stated pipeline Unknown
  • Company target Unknown
GPUs operational
Unknown
GPUs contracted
Unknown
GPUs announced
Unknown

GPU platforms: NVIDIA B200, GB200, GB300

Financial position

Annualised revenuereported run-rate
$1.1B
Revenue growth (YoY)
Unknown
Backlog
Unknown
Disclosed contract value
Unknown
Valuationpost money funding valuation
$8.3B
Equity raised (captured)
$800M
Debt raised (captured)
Unknown
Total debt (reported)
Unknown
Hyperscaler credit supportnot cash raised
Unknown

Scorecard

score-v1.0 · 0–100 · scores with under 34% evidence coverage stay empty

Business Model Quality 48

evidence coverage 75%

  • contracted revenue model: 45 (weight 0.4) — spot/on-demand oriented
  • model diversity: 66 (weight 0.2) — inference_as_a_service, dedicated_clusters, ai_platform_software
  • margin: unknown (weight 0.25) — no margin data
  • software upside: 30 (weight 0.15) — 2 software layers
Capital Access 59

evidence coverage 90%

  • total capital raised: 47 (weight 0.3) — $800M equity+debt captured
  • raised last 12m: 50 (weight 0.25) — $800M in 12m
  • valuation: 64 (weight 0.2) — post_money_funding_valuation valuation $8.30B
  • strategic backers: 90 (weight 0.15) — NVIDIA/AMD investor or hyperscaler credit support
  • cost of debt: unknown (weight 0.1) — no rate disclosed
Competitive Moat 43

evidence coverage 75%

  • physical control: 30 (weight 0.3) — power/land/DC control
  • software depth: 32 (weight 0.25) — 2 software layers
  • contracted lock in: unknown (weight 0.25) — backlog/contract value UNKNOWN
  • operational reputation: 75 (weight 0.2) — Gold (ClusterMAX 2.0)
Concentration Risk (higher = riskier) Insufficient evidence

evidence coverage 0%

  • top customer share: unknown (weight 0.8) — UNKNOWN
  • few customers: unknown (weight 0.2) — 0 disclosed customers
Customer Diversification Insufficient evidence

evidence coverage 0%

  • top customer share: unknown (weight 0.6) — concentration UNKNOWN
  • distinct customers: unknown (weight 0.4) — 0 disclosed customer relationships
Customer Quality Insufficient evidence

evidence coverage 0%

  • hyperscaler or lab: unknown (weight 0.5) — 0 disclosed customers
  • contracted value: unknown (weight 0.5) — disclosed contract value UNKNOWN
Execution Risk (higher = riskier) Insufficient evidence

evidence coverage 0%

  • build to operating ratio: unknown (weight 0.45) — 0 MW to build vs 0 MW operating
  • soft pipeline share: unknown (weight 0.25) — 0 MW secured/announced/speculative
  • organisational youth: unknown (weight 0.3) — founded UNKNOWN
Financial Durability 56

evidence coverage 40%

  • revenue scale: 62 (weight 0.3) — annualised revenue $1.10B (reported run-rate)
  • net margin: unknown (weight 0.2) — no margin data
  • cash: unknown (weight 0.2) — cash UNKNOWN
  • backlog coverage: unknown (weight 0.2) — no backlog/revenue
  • public disclosure: 40 (weight 0.1) — private
Financial Risk (higher = riskier) 27

evidence coverage 40%

  • losses: unknown (weight 0.3) — no margin data
  • debt to revenue: unknown (weight 0.3) — no debt/revenue data
  • reliance on new capital: 20 (weight 0.25) — 12m capital raised / revenue 0.7x
  • revenue scale inverse: 38 (weight 0.15) — revenue $1.10B
Geographic Position 29

evidence coverage 100%

  • countries: 18 (weight 0.6) — United States
  • sovereign positioning: 45 (weight 0.4) — no sovereign positioning
GPU Supply Position 93

evidence coverage 60%

  • gpu fleet and contracted: unknown (weight 0.4) — 0 GPUs operational+contracted
  • vendor relationship: 95 (weight 0.3) — NVIDIA investor/Exemplar
  • latest generation: 90 (weight 0.3) — NVIDIA B200, GB200, GB300
Growth Momentum 33

evidence coverage 35%

  • revenue growth yoy: unknown (weight 0.4) — no growth data
  • capacity momentum: unknown (weight 0.25) — 0 MW contracted/under construction
  • recent major events: 25 (weight 0.2) — 1 major confirmed events in 180d
  • capital raised 12m: 43 (weight 0.15) — raised $800M in 12m
Infrastructure Execution 53

evidence coverage 60%

  • delivered projects: 50 (weight 0.4) — 1/2 tracked projects (partially) operational
  • operational share of hard mw: unknown (weight 0.4) — 0 of 0 hard MW operational
  • building now: 60 (weight 0.2) — 1 under construction
Management Execution 35

evidence coverage 50%

  • delivered capacity: unknown (weight 0.5) — 0 MW delivered to date
  • milestones 180d: 25 (weight 0.3) — 1 confirmed major milestones (180d)
  • disclosure discipline: 50 (weight 0.2) — private
Operational Capacity Insufficient evidence

evidence coverage 0%

  • operational mw: unknown (weight 0.65) — operational MW UNKNOWN
  • operational gpus: unknown (weight 0.35) — operational GPUs UNKNOWN
Power Pipeline Quality Insufficient evidence

evidence coverage 20%

  • weighted pipeline mw: unknown (weight 0.5) — ladder-weighted 0 MW
  • hard share: unknown (weight 0.3) — no MW data
  • physical power control: 30 (weight 0.2) — physical-layer control 30/100
Software Platform Maturity 60

evidence coverage 100%

  • software layers: 32 (weight 0.5) — inference, training
  • software control: 100 (weight 0.25) — software-layer control 100/100
  • independent rating: 75 (weight 0.25) — Gold (ClusterMAX 2.0)
Strategic Partnerships 93

evidence coverage 60%

  • vendor equity: 95 (weight 0.4) — NVIDIA investor
  • hyperscaler ties: unknown (weight 0.4) — none evidenced
  • exemplar or rating: 90 (weight 0.2) — NVIDIA Exemplar / top ClusterMAX
Technical Differentiation 23

evidence coverage 100%

  • differentiators: 0 (weight 0.7) — no clear differentiator evidenced
  • independent rating: 75 (weight 0.3) — Gold (ClusterMAX 2.0)
Technical Strength 65

evidence coverage 70%

  • independent rating: 75 (weight 0.45) — Gold (ClusterMAX 2.0)
  • exemplar cloud: unknown (weight 0.15) — n/a
  • fabric evidence: unknown (weight 0.15) — fabric UNKNOWN
  • stack documentation: 48 (weight 0.25) — 4 technology layers evidenced
Vertical Integration 72

evidence coverage 100%

  • stack control: 72 (weight 1) — 10 of 12 layers evidenced; 2 UNKNOWN
Career Opportunity 57

evidence coverage 83%

  • growth momentum: 33 (weight 0.2) — Growth Momentum
  • capital access: 59 (weight 0.15) — Capital Access
  • technical strength: 65 (weight 0.12) — Technical Strength
  • software platform maturity: 60 (weight 0.12) — Software Platform Maturity
  • power pipeline quality: unknown (weight 0.1) — Power Pipeline Quality
  • geographic position: 29 (weight 0.06) — Geographic Position
  • strategic partnerships: 93 (weight 0.08) — Strategic Partnerships
  • hiring: unknown (weight 0.07) — open roles (public boards)
  • inv financial risk: 73 (weight 0.1) — 100 - financial risk
Data Quality 60

evidence coverage 47%

  • dq coverage: unknown (weight 0) — 0.471
  • dq avg confidence: unknown (weight 0) — 0.732
  • dq primary source pct: unknown (weight 0) — 0.375
  • dq freshness days: unknown (weight 0) — 0
  • dq conflicts: unknown (weight 0) — 0
  • dq missing count: unknown (weight 0) — 9
Worth Investigating? (not investment advice) 51

evidence coverage 68%

  • growth momentum: 33 (weight 0.18) — Growth Momentum
  • technical differentiation: 23 (weight 0.12) — Technical Differentiation
  • customer quality: unknown (weight 0.12) — Customer Quality
  • power pipeline quality: unknown (weight 0.12) — Power Pipeline Quality
  • capital access: 59 (weight 0.12) — Capital Access
  • strategic partnerships: 93 (weight 0.1) — Strategic Partnerships
  • technical strength: 65 (weight 0.08) — Technical Strength
  • infrastructure execution: 53 (weight 0.08) — Infrastructure Execution
  • operational capacity: unknown (weight 0.08) — Operational Capacity
  • novelty bonus: 0 (weight 0.3) — company ≤3 years old
Overall Scale Potential Insufficient evidence

evidence coverage 35%

  • power pipeline quality: unknown (weight 0.3)
  • capital access: 59 (weight 0.25)
  • customer quality: unknown (weight 0.2)
  • operational capacity: unknown (weight 0.15)
  • vertical integration: 72 (weight 0.1)
Overall Survival Probability (5y, %) 86

evidence coverage 57%

  • model: unknown (weight 1) — logistic(durability, capital access, customer quality, risks)

Forecasts (rule-based, P10–P90)

HorizonOperational MWAnnualised revenue
12 months Unknown $1.4B ($1.2B–$1.5B)
24 months Unknown $1.8B ($1.4B–$2.1B)
36 months Unknown $2.2B ($1.5B–$2.7B)
60 months Unknown $3.2B ($1.8B–$4.4B)

Stored in an immutable prediction ledger and scored against reported outcomes when they fall due. Predictions, not facts.

Outcome probabilities

Outcome12 mo24 mo36 mo60 mo
Remain independent91%82%73%58%
Acquired6%12%17%27%
Merge1%2%3%5%
Fail2%5%7%11%
IPO14%26%36%53%
Acquire another company4%8%12%18%
Financial distress1%2%3%5%
Major cloud scale (≥2 GW operational or ≥$10B revenue)0%0%0%0%
Remain niche98%95%93%89%
Strategic pivot5%10%15%25%
Exit AI infrastructure1%2%3%5%

Ownership outcomes (independent, acquired, merge, fail) sum to 100% at each horizon; the others are independent events.

Vertical-integration stack

  • energy THIRD PARTY Partner facilities. (inferred)
  • land THIRD PARTY Partner facilities. (inferred)
  • data center PARTNERED 5C AI factories. (inferred)
  • power cooling PARTNERED 5C. (inferred)
  • gpu hardware OWNED Operates GPU clusters. (inferred)
  • network UNKNOWN Not captured. (inferred)
  • storage UNKNOWN Not captured. (inferred)
  • bare metal OWNED GPU clusters. (inferred)
  • orchestration OWNED Own platform. (inferred)
  • cloud api OWNED APIs. (inferred)
  • ai platform OWNED Inference/fine-tuning platform. (inferred)
  • managed ai OWNED Managed inference. (inferred)

Not yet sourced: power mw.operational, power mw.contracted, gpu count, cash, backlog, employees, profile.ceo, profile.founded year, customers.

Timeline

  1. 1 Jul 2026 · equity

    Together AI — Series C

    Source: TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation
  2. 20 Feb 2025 · equity

    Together AI — Series B $3.3B valuation 17 months before Series C (per TechCrunch); amount not captured.

    Source: TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation

Recent coverage

Updated 1 Nov 2026, 11:54 UTC

  1. 1 Oct 2026 NeoCloud

    TPC Fall 2026 Hackathon at LRZ Brings Together AI for Science Community

    Oct. 1, 2026 — From Sept. 9-11, 2026, the Leibniz Supercomputing Centre (LRZ), hosted the autumn hackathon of the Trillion Paramter Consortium. Organized by the EuroTPC, the hackathon gathered more than 30 researchers from leading European and US supercomputing centers and universities to turn ideas into practical…

    HPCwire