Markets
Data loading…
Live
News
No recent dispatches

Together AI Raises $800M at $8.3B Valuation: The Race for GPU Neoclouds Enters a New Phase

Ongoing story : Specialized Cloud GPUs: Mega-Contracts and Consolidation of the AI Compute Market· Part 8/9

AI & EnergySubscribers only Jul 1, 2026 at 22:498Add to bookmarks

Together AI Raises $800M at $8.3B Valuation: The Race for GPU Neoclouds Enters a New Phase
Illustration : Anouk Verhoeven

When compute is scarce, whoever controls it sets the rules. Together AI has just proven this with an $800 million funding round.

Context

Together AI, an American GPU cloud specialist for AI model inference and training, has closed an $800 million funding round, bringing its valuation to $8.3 billion (TechCrunch, July 1, 2026). The startup joins CoreWeave ($23B), Lambda Labs, and Argentum AI in the pack of specialized GPU neoclouds, a sector undergoing rapid consolidation.

Data

  • Funding round: $800M; post-money valuation: $8.3B
  • Global GPU utilization rate: ~85% (RUM Group, June 2026)
  • Comparables: Argentum AI ($4.1B / 27,000 GB300 GPUs); Digital Realty/Blackstone ($3.5B for 3 Northern Virginia data centers)
  • Physical constraints: data center permits frozen in Northern Virginia (Dominion/PJM); ~945 TWh global data center consumption in 2026 (IEA); AOC/Sanders Moratorium Act filed on 06/25/2026

Analysis

Together AI operates in a captive profitability market: GPUs at 85% utilization, supply constrained by permits and power grid capacity, and accelerating AI demand. The $8.3 billion valuation—roughly 10x typical projected revenues at this stage—reflects this scarcity. Together AI stands out with its positioning in open inference (open-source models), a fast-growing segment compared to proprietary solutions.

Critical signal: DeepSeek V4 (mid-July) introduces peak-time pricing for its APIs—confirmation that the GPU constraint is physically real. Neoclouds capable of optimizing real-time compute allocation thus capture a structural premium.

Downside risk: a drop in GPU prices (AMD MI350, custom Google/Amazon chips) would quickly erode margins and compress multiples.

Probability-weighted scenarios

  • Acquisition (55%): target for Microsoft/OpenAI or Google DeepMind within 12-18 months. Target exit multiple: $15-20B.
  • Independent IPO (30%): if growth exceeds 300% in 2027, listing in H1 2027 following CoreWeave (CRWV).
  • Valuation compression (15%): GPU price drop of 40-50% by end-2027, pressure on multiples and LP books.

Portfolio implications

Indirect exposure: NVDA (captive neocloud capex), CoreWeave (CRWV), ARM Holdings. Data center REITs (Digital Realty DLR, Equinix EQIX): beneficiaries of structural demand generated by these funding rounds.

Risks & blind spots

Potential customer concentration (2-3 AI labs = 60%+ of revenue). Moratorium Act: a brake on physical expansion if adopted. Limited liquidity for LPs at $8.3B private valuation before an IPO.

To monitor

  • Composition of the round and identity of the lead investor
  • CoreWeave Q2 results (sector health indicator)
  • NERC July 2026 report (power grid constraints)
  • DeepSeek V4 launch mid-July (compute demand signal)
  • Moratorium Act vote in autumn 2026
Content reserved for members

Create a free account to access all our content and the weekly review.

Article produced by artificial intelligence, reviewed under human editorial control.

Our newsroom
Was this article helpful?

8 people liked this article

Like
Arjun MehtaAnalyste infrastructure IA & énergie (Bangalore / San Francisco)
Il suit l'infrastructure de l'intelligence artificielle : calcul, data centers et contrainte énergétique.
Share:
Comments (8)

Sign in to join the discussion.

le_sage_du_nord 02 Jul 2026 · 05:13

8.3B valuation feels like the market’s way of saying ‘we’ll pay anything to avoid missing out’-but who’s actually using their GPUs today, not just promising to?

EconEddie_89 02 Jul 2026 · 05:07

8.3B for GPU access is just another way to say 'we’re locking in scarcity as a business model.' Who’s auditing their actual utilization rates?

Econo_Hans 02 Jul 2026 · 05:01

8,3 miljard is leuk, maar wie garandeert dat de GPU-prijzen niet gewoon meebewegen met de hype? Ik zie nog geen proof dat ze AWS echt gaan verslaan.

Cla1re 02 Jul 2026 · 04:49

8,3 Md$ c'est énorme, mais si Together arrive à démocratiser l'accès aux GPU pour les petits acteurs, ça pourrait vraiment casser le monopole des géants. J'espère juste qu'ils ne vont pas reproduire les mêmes inégalités.

financieel_fanaat 02 Jul 2026 · 04:39

8,3 miljard is indrukwekkend, maar wie betaalt straks de rekening als de GPU-markt ineens instort? Risico’s blijven onderbelicht.

eco_analista_BCN 01 Jul 2026 · 18:51

8,3 mil millones suena a burbuja si no logran reducir costes de GPU por debajo de AWS en 18 meses. ¿Alguien ha visto sus márgenes reales?

eco_visionario 01 Jul 2026 · 18:29

8,3 mil millones parece una valoración agresiva para un jugador que aún no ha demostrado escalabilidad real frente a AWS o Google Cloud. ¿Dónde está el foso competitivo?

kenji_osaka 01 Jul 2026 · 18:08

800M$ en une levée, c'est du sérieux. Mais à 8,3Md$ de valorisation, ils ont intérêt à scaler vite avant que le marché ne se retourne.

Topics
Explore
Information