The Venture Capital Podcast with Fexingo · 2026-07-01 · 7 min
Key moments - from our scoring
Substance score
54 / 100
Five dimensions, 20 points each
Together AI's $800 million Series B signals a major venture capital pivot toward AI infrastructure over model development. The company operates as a GPU cloud provider - essentially renting NVIDIA H100 clusters to AI startups and enterprises that need compute without building their own infrastructure or relying on hyperscalers. Led by General Catalyst and Accel with sovereign wealth participation, the round values the company at $8.3 billion, implying ~$800 million in expected annual revenue. Together AI's competitive advantage centers on offering a tailored AI stack using open-source software with direct NVIDIA hardware access and flexible pricing, positioning them as a neutral multi-cloud provider against AWS, Google Cloud, and Azure. The valuation reflects investor conviction that GPU compute demand will remain insatiable as foundation model companies scale, but carries material execution risk: Together AI must deploy capital efficiently into NVIDIA hardware, maintain high utilization rates above 70%, and lock in long-term contracts before commoditization. This represents a fundamental shift in VC capital allocation toward capital-intensive infrastructure plays with clear unit economics ($30,000 GPU generating $3/hour rental revenue) rather than software-first startups.
Together AI rents GPU compute capacity (specifically NVIDIA H100 clusters) to AI startups and enterprises. They buy GPUs for approximately $30,000 each and rent them for $3 per hour, generating revenue by maintaining high utilization rates above 70%.
The round was led by late-stage crossover funds including General Catalyst and Accel, with additional participation from sovereign wealth funds.
Together AI offers a more AI-specific stack with open-source software, direct access to NVIDIA's latest hardware, and flexible pricing versus the hyperscalers' general-purpose optimization, while positioning as a neutral multi-cloud provider to prevent customer lock-in.
The valuation implies approximately $800 million in expected annual revenue, representing roughly 10x forward revenue multiple for an infrastructure business.
The primary risk is capex intensity and demand sustainability - if AI investment cools, the company is left with expensive NVIDIA hardware and underutilized capacity with limited alternative uses, making it a stranded asset problem.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode provides a solid overview of Together AI's market position and the broader infrastructure thesis, including unit economics (buy GPU for $30k, rent at $3/hour, need 70% utilization) and the multi-cloud positioning argument. However, much of the discussion rehashes predictable venture patterns (commoditization will happen, VCs seek board seats, hyperscalers are a threat) without probing deeper into execution risks, unit economics validation, or market size constraints.
you buy a GPU for $30,000, you rent it out for $3 an hour, and if you keep utilization above 70%, you make a good return
Startups don't want to be locked into one hyperscaler. They want a neutral provider. Together AI offers that.
The conversation hits familiar venture talking points - multi-cloud trends, infrastructure beats software, the unit economics narrative - without much fresh analysis. The observation that VCs are moving from software to capex-heavy hardware is accurate but well-worn by mid-2024. No contrarian takes, first-principles critiques, or surprising counterarguments emerge.
The thesis is straightforward: Together AI sells GPU cloud access, essentially renting out NVIDIA H100 clusters to AI startups and enterprises.
Investors want to see clear revenue paths. Infrastructure, on the other hand, has a very clear unit economics
Lucas appears to be a host or analyst offering informed commentary on the round, but no operator or practitioner with direct experience at Together AI, a GPU cloud competitor, or a major customer is present. The discussion lacks the credibility that would come from someone who has actually built or managed compute infrastructure at scale or negotiated contracts with Together AI.
Lucas: So Together AI just announced an $800 million Series B
I've heard they have commitments from a few of the big foundation model companies
The episode includes some concrete data - $800M round size, $8.3B post-money valuation, 10x forward revenue multiple, $800M assumed revenue, $30k GPU cost, $3/hour rental rate, 70% utilization threshold, NVIDIA stock movements, ARK fund performance. However, the claimed revenue figure ($800M) is presented as assumption, not confirmed fact, and key claims about multi-year contracts and customer names are vague ('a few of the big foundation model companies').
At $8.3 billion post-money, investors are paying roughly 10x forward revenue, assuming they're on track for around $800 million in revenue this year
you buy a GPU for $30,000, you rent it out for $3 an hour, and if you keep utilization above 70%, you make a good return
Luna and Lucas maintain a smooth back-and-forth with logical follow-ups (e.g., Luna asking about competition with hyperscalers, then asking about risks to AI model companies). However, questions lack edge - no genuine pushback on the $800M revenue assumption, no skepticism about the ability to sustain 70% GPU utilization, no deep drilling on the term sheet claims, and no challenge to cheerleading about 'the dream' of becoming 'the AWS of AI.'
But doesn't that put them in direct competition with the hyperscalers? AWS, Google Cloud, Azure?
If the AI hype cycle cools, these capex-heavy businesses could be left with expensive hardware and no customers.
Computed from the transcript - who did the talking, and the words that came up most.
In this episode of The Venture Capital Podcast, Lucas and Luna dive into Together AI's massive $800 million Series B, valued at $8.3 billion. They explore what this means for the AI infrastructure market, how VCs are thinking about the shift from model companies to compute providers, and why this round signals a new era of capital concentration in AI. Lucas breaks down the economics: the $800 million raise gives Together AI a post-money valuation of $8.3 billion, implying investors are paying roughly 10x forward revenue for a company that sells GPU cloud access. They contrast this with the struggling model-only startups and discuss the 'GPU-as-a-service' thesis that's driving VC dollars. Luna questions whether this is sustainable, especially given the capex demands. They also touch on how this affects earlier-stage AI startups trying to raise. A must-listen for anyone following AI venture capital in 2026.
Transcribed and scored by The B2B Podcast Index.
Lucas: So Together AI just announced an $800 million Series B, valuing them at $8.3 billion. And I think this is one of the most telling venture rounds we've seen all year. Luna: Eight hundred million - that's a huge check.
Who led it, and what's the thesis here? Lucas: The round was led by a mix of late-stage crossover funds - think firms like General Catalyst and Accel, plus some sovereign wealth money. The thesis is straightforward: Together AI sells GPU cloud access, essentially renting out NVIDIA H100 clusters to AI startups and enterprises. They're positioning themselves as the compute layer for the AI stack, not an AI model company.
Luna: Right, so they're an infrastructure play, not a pure AI startup. That distinction seems to matter more and more. Lucas: Exactly. And the valuation tells the story.
At $8.3 billion post-money, investors are paying roughly 10x forward revenue, assuming they're on track for around $800 million in revenue this year. That's a premium for an infrastructure business, but VCs are betting that the demand for compute is insatiable. Luna: But $800 million in revenue - that's a lot of GPU hours.
How do they justify that kind of scale? Lucas: They've been aggressive in signing multi-year contracts with large AI labs. I've heard they have commitments from a few of the big foundation model companies - the ones that haven't built their own compute clusters yet. And they're also selling to enterprises that want to fine-tune models on sensitive data without going through a public cloud.
Luna: So it's almost like a private cloud for AI. But doesn't that put them in direct competition with the hyperscalers? AWS, Google Cloud, Azure? Lucas: It does, but the argument is that the hyperscalers are optimized for general-purpose workloads, not specifically for AI training and inference.
Together AI offers a more tailored stack - they use open-source software, they have direct access to NVIDIA's latest hardware, and they're more flexible on pricing and contracts. Luna: That makes sense. And this round definitely signals that VCs see the AI infrastructure market as the next big wave. We saw earlier this year with the CoreWeave deals, and now this.
Lucas: Right. And look at the market data today - NVIDIA's stock is up modestly, but the ARK Genomic Revolution ETF, which has a lot of AI exposure, is up over 5% in the last five days. The market is pricing in continued AI investment. Luna: But I'm curious - what does this mean for the AI model companies themselves?
If investors are piling into infrastructure, are they abandoning model startups? Lucas: It's a shift. The model companies that have struggled to monetize - like Stability AI or some of the open-source players - are finding it harder to raise. Investors want to see clear revenue paths.
Infrastructure, on the other hand, has a very clear unit economics: you buy a GPU for $30,000, you rent it out for $3 an hour, and if you keep utilization above 70%, you make a good return. Luna: So the risk is more about demand sustainability. If the AI hype cycle cools, these capex-heavy businesses could be left with expensive hardware and no customers. Lucas: That's the bear case.
And it's not trivial - Together AI likely has to spend a significant portion of that $800 million on buying GPUs from NVIDIA. They're essentially converting equity into hardware. If demand softens, they're stuck. Luna: But VCs seem to be betting that AI compute demand will keep growing for years.
And the numbers today support that - look at the mega-caps: Microsoft up nearly 10% in five days, Alphabet up almost 5%. They're all spending billions on AI. Lucas: Exactly. And this is also a bet on the multi-cloud trend.
Startups don't want to be locked into one hyperscaler. They want a neutral provider. Together AI offers that. Luna: So for a VC, the question becomes: is this a winner-take-all market?
Or can multiple GPU cloud providers coexist? Lucas: I think it's more of a land-grab right now. There's enough demand for a few big players. But eventually, commoditization will set in.
The key is to lock in long-term contracts before that happens. Luna: Speaking of long-term thinking - we should mention that shows like this one are ad-free and listener-supported. If these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going. Lucas: Yeah, it's at buy me a coffee dot com slash fexingo.
That's 'buy me a coffee' spelled out, dot com slash fexingo. It makes a real difference in keeping us independent. Luna: And we really appreciate it. Now, back to Together AI - one thing I want to dig into is the term sheet.
What kind of terms did they get at that valuation? Lucas: Good question. At $8.3 billion, it's likely a standard late-stage structure: participating preferred, 1x liquidation preference, maybe some anti-dilution protection.
But the interesting part is the governance - I'd bet the lead investors got a board seat and some veto rights over future capital allocation, given the capex intensity. Luna: So the VCs are taking an active role in how the company spends its capital. That makes sense when the business model is so hardware-heavy. Lucas: Right.
And this also shows how the venture model is evolving. Traditionally, VCs invested in software with high margins and low capital needs. Now they're backing businesses that require billions in infrastructure. It's a different risk profile.
Luna: But the returns could be enormous if it works. If Together AI becomes the AWS of AI, we're talking about a multi hundred billion dollar company. Lucas: That's the dream. But the path is narrow.
They need to execute flawlessly on deployment, keep utilization high, and fend off competition from both hyperscalers and other GPU cloud startups. Luna: It's a high-stakes bet. But that's what venture capital is all about. Lucas: Exactly.
And this round will be a case study for years to come on how to raise massive capital for an infrastructure-heavy AI business. Luna: Well, we'll be watching. Thanks, Lucas. Lucas: Thanks, Luna.
And thanks to our listeners for joining us.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.
Radiology Can't Keep Up. Here's Where AI Actually Helps | Dr. Nina KottlerRethink Imaging · on Foundation models96 / 100
Joe Yaffe, COO at Cowboy Space, on Whether Data Centers in Space Are CheaperBetween Two COO's with Michael Koenig · on AI infrastructure90 / 100
The Best of Series Six: The Professional Investment PodcastThe Professional Investment Podcast · on Venture capital81 / 100
The $500 Billion AI Debt Machine & The Great CapEx TestMarket Maker · on CoreWeave80 / 100
John Suarez on How Businesses Should Actually Use AI Without Losing Human JudgementMarketing for SMEs · on Foundation models78 / 100
MicroConf Tactics: Start a SaaS From $0 in 2026MicroConf On Air · on AWS77 / 100
How VCs Are Pricing Startup Burn Rate in 202652 / 100Why VCs Are Underwriting Startup Supply Chain ResilienceHow VCs Are Pricing Startup Exits in 2026How VCs Are Valuing Startup Community in 2026How VCs Are Pricing Startup Adaptability in 2026