Together AI Success Story: How an Open-Source AI Startup Became an $8.3 Billion Infrastructure Giant

Every gold rush needs someone selling shovels. During the California gold rush of the 1800s, the people who got rich weren't always the ones panning for gold. Many of them were selling the tools, the tents, and the transportation that made the search possible.

The AI boom is following the same script, just with GPUs instead of pickaxes.

While OpenAI, Anthropic, and Google fight over who can build the smartest model, a quieter battle is happening underneath them. It's a battle over compute, infrastructure, and the plumbing that makes any of this AI magic actually run. And some of the biggest winners of this decade may not be the model makers at all. They might be the infrastructure builders.

Nvidia is the most obvious example. It doesn't build chatbots. It builds the chips everyone else needs to build chatbots, and that single decision made it one of the most valuable companies on Earth.

But Nvidia isn't alone anymore. A newer category of company, sometimes called a "neocloud," has emerged to rent out GPU power, run AI models at scale, and give developers a faster, cheaper alternative to the giant hyperscalers. CoreWeave rode this wave to a public listing. Lambda has carved out its own lane. And then there's Together AI, a company that took a slightly different bet than most of its rivals.

Instead of building its own proprietary model to compete with GPT or Claude, Together AI decided to bet everything on open-source AI. It built the cloud, the tooling, and the performance engineering that let anyone run open models like Llama or DeepSeek cheaply, quickly, and at serious scale.

That bet just got a massive vote of confidence. In July 2026, Together AI raised $800 million at an $8.3 billion valuation, according to TechCrunch. That's more than double the $3.3 billion valuation it held just seventeen months earlier. This is the story of how that happened.

What Is Together AI?

Together AI is an AI infrastructure company, sometimes called an "AI Acceleration Cloud" or "AI-native cloud," founded in June 2022 and headquartered in San Francisco, California.

In plain terms, Together AI does not build its own foundation model to rival GPT-5 or Claude. Instead, it builds the cloud platform that lets developers and enterprises train, fine-tune, and run other companies' open-source AI models, quickly and at a fraction of the cost of closed, proprietary systems.

Think of it this way. If OpenAI and Anthropic are like car manufacturers building their own vehicles from scratch, Together AI is more like the highway system, the gas stations, and the repair shops that let any car, built by anyone, actually go somewhere useful.

The company's core offerings fall into a few buckets:

• GPU Cloud — renting out Nvidia GPU capacity, from single nodes to massive multi-thousand-GPU clusters
• Inference — running trained models so applications can generate responses in real time
• Training and fine-tuning — helping companies customize open models on their own data
• Enterprise AI infrastructure — dedicated clusters, security, and deployment tools for larger organizations

Together AI's founders describe the mission simply: they believe the future of AI shouldn't belong to two or three closed labs. They want a world where open models can match proprietary ones on quality while costing a fraction as much to run.

That mission has turned into a real business. The company says its platform now serves over a million developers, and its annual bookings have topped $1.15 billion, according to the company's own disclosures reported by TechCrunch around its Series C announcement.

Meet the Founders

Together AI wasn't built by first-time founders chasing a trend. It was built by people who had already spent years, in some cases decades, working on the hard problems sitting underneath modern AI.

Vipul Ved Prakash, the CEO, is the connective tissue of the group. Born in New Delhi, he studied mathematics, physics, and computer science at St. Stephen's College in Delhi before dropping into software development. He co-founded Cloudmark, a cybersecurity company focused on email security that was later acquired by Proofpoint. Then he built Topsy, a social media search and analytics company, which he sold to Apple in 2013 for a reported price north of $200 million, according to TechCrunch. Inside Apple, Prakash worked on Siri search and other AI-related projects, giving him a front-row seat to how a tech giant approached large-scale AI years before ChatGPT existed.

Ce Zhang, the company's CTO, brought the academic rigor. He earned a math degree from Peking University and a PhD from the University of Wisconsin-Madison, later teaching computer science at ETH Zurich and the University of Chicago. His research focus, making machine learning cheaper and more accessible, reads almost like a mission statement for the company he'd go on to co-found.

Chris Ré, a Stanford professor, is one of the most respected names in AI systems research. He has a track record of turning academic research into real companies, and his lab's work on data systems and machine learning efficiency has influenced much of the modern AI infrastructure stack.

Percy Liang, also a Stanford professor, is a leading voice in AI research and evaluation, known for pushing the field toward more rigorous, transparent benchmarking of language models.

Tri Dao joined as Chief Scientist, and his name might be the most recognizable to working AI engineers. Dao created FlashAttention, an open-source technique that dramatically speeds up how transformer models handle memory and computation. It's since been adopted by OpenAI, Anthropic, Meta, and Mistral to train their own frontier models. Having the creator of one of the most widely used efficiency tricks in modern AI sitting inside your own company is not a small advantage.

According to Prakash, the idea for Together AI crystallized when the four core founders realized foundation models represented a shift on the scale of the invention of the transistor. They watched a small number of well-funded corporations start to control access to that shift, purely because training and running these models required GPU clusters that cost more than most companies could ever afford. Meanwhile, the open research community that had powered AI progress for a decade had almost no seat at the table.

That tension, powerful models locked behind a few walled gardens, is what Together AI was built to break open.

The Problem They Wanted to Solve

To understand why Together AI exists, it helps to picture the AI landscape around 2022, right as the founders were getting started.

GPUs were scarce. Nvidia's most powerful chips were backordered for months, and only companies with deep pockets or direct relationships could get meaningful allocations. Cloud computing giants controlled most of that supply, and they weren't cheap.

Training or even running a large language model required serious infrastructure expertise, the kind that most startups simply didn't have in-house. Even when a company managed to get GPU access, actually squeezing good performance out of that hardware was its own specialized skill.

At the same time, the most capable AI systems lived behind closed doors. If you wanted frontier-level performance, you typically had to use a proprietary API from a handful of labs, at their price, under their terms, with limited visibility into how your data was handled.

Open-source models existed, but running them well, quickly, and affordably was still genuinely hard. Inference was slow. Deployment was clunky. And most developers didn't have the systems expertise to optimize a model's performance the way a research lab could.

Together AI's founders saw all of these problems as one problem: there was no serious infrastructure layer built specifically for open AI. So they decided to build it themselves.

Building Together AI

The company launched in June 2022, technically incorporated as Together Computer Inc. The timing turned out to be extraordinary. Just months later, ChatGPT launched and the entire world suddenly cared about generative AI.

Early funding validated the founders' pedigree more than any product traction could have. The seed round, roughly $20 million, was led by Lux Capital, with a long list of backers including Factory, SV Angel, First Round Capital, Long Journey Ventures, and individual angels like Cloudera co-founder Jeff Hammerbacher and Oasis Labs founder Dawn Song.

The company's early technical work centered on making open-source models genuinely competitive on performance, not just cost. Tri Dao's FlashAttention research became a foundation for the company's inference stack, and the team built out techniques with names like Medusa and Flash-Decoding that squeezed more speed out of the same hardware.

By late 2023, the pitch had sharpened: Together AI wasn't just a place to rent GPUs. It was building what it called the fastest inference stack for transformer models, giving developers quick access to more than 100 open models through a single API. That story helped the company close a $102.5 million Series A in November 2023, led by Kleiner Perkins, with participation from Nvidia and Emergence Capital, a round profiled in detail by Newcomer.

Credibility built quickly after that. Enterprise names like Salesforce and Zoom began showing up as customers. The company kept shipping research, kept optimizing its inference stack, and kept expanding its GPU footprint, all while the broader AI industry was realizing that inference costs, not just training costs, were about to become the real battleground.

That credibility translated directly into capacity. By November 2024, Together AI had lined up a partnership with Hypertec to co-build a cluster of 36,000 Nvidia GB200 NVL72 GPUs, essentially a rack-scale supercomputer design that links 72 GPUs together over Nvidia's high-speed NVLink connection, giving the company the raw hardware it needed to court larger training and inference workloads. That buildout set up the next fundraising milestone: a $305 million Series B in February 2025, which arrived alongside the company's own Blackwell-powered GPU clusters. From there, the growth compounded quickly. Through the rest of 2025, Together AI kept extending its footprint into new markets and new product categories, groundwork that would carry straight through to its most recent, and largest, funding round.

The Business Model

Together AI makes money the way most modern infrastructure companies do: through usage, not just subscriptions. But it's worth breaking apart the layers, because each one targets a different kind of customer.

GPU Cloud is the foundation. Customers rent Nvidia GPU capacity, everything from H100 and A100 chips to the newer Blackwell generation, priced by the hour or by usage. This is for teams that want raw compute without owning physical hardware.

Inference API is arguably the company's signature product. Rather than running their own servers, developers send requests to Together's API and get responses back from any of more than 200 supported open-source models, spanning chat, image, audio, vision, code, and embeddings. Pricing here is typically usage-based, tied to tokens processed.

Fine-tuning lets companies customize an open model on their own data without building training infrastructure from scratch. This turns a generic open model into something tailored to a specific business.

Dedicated clusters, sometimes marketed as Together Reasoning Clusters or Together GPU Clusters, serve companies with heavy, predictable workloads that want guaranteed capacity rather than shared, variable access.

Enterprise deployments bundle all of the above with security, compliance, and support commitments that larger organizations require before they'll trust a vendor with production workloads.

The pricing model blends usage-based billing for smaller developers with negotiated enterprise contracts for the largest customers, the kind that commit to spending millions of dollars a year in exchange for guaranteed capacity and dedicated support. This mix is why the company talks about "annual bookings" rather than just revenue. Bookings represent contracted future spend, while revenue reflects money already recognized.

Funding Journey

Together AI's capital raising has followed the classic AI-era pattern: bigger rounds, faster, at valuations that would have seemed absurd just a few years earlier.

Investor participation across these rounds has included Nvidia, Kleiner Perkins, Emergence Capital, Prosperity7 Ventures, Salesforce Ventures, General Catalyst, Coatue, March Capital, Lux Capital, DAMAC Capital, Vista Equity Partners, Pegatron, and SentinelOne's S Ventures, according to disclosures from Together AI's own Series B announcement, PR Newswire, and TechCrunch.

According to figures compiled by Sacra and Crunchbase, Together AI had raised roughly $533.5 million across its first four rounds before the Series C. Adding the $800 million Series C brings total disclosed funding to well over $1.3 billion, though some outlets, including Tech Funding News, have cited a slightly lower cumulative figure of "more than $1.1 billion." Where sources differ on cumulative totals, this article treats the individually confirmed round amounts as the reliable baseline.

Notably, the March 2026 report from The Information that first surfaced talk of a new mega-round pegged the target valuation at $7.5 billion. The company ended up closing at $8.3 billion instead, which suggests investor demand for the round outpaced even the ambitious early expectations.

Timeline of Major Milestones

• June 2022 — Together Computer Inc. is founded by Vipul Ved Prakash, Ce Zhang, Chris Ré, and Percy Liang, with Tri Dao joining as founding Chief Scientist.
• 2023 — Seed round of roughly $20 million closes, led by Lux Capital.
• November 2023 — $102.5 million Series A closes, led by Kleiner Perkins, with Nvidia and Emergence Capital participating.
• March 2024 — $106 million round closes at a $1.25 billion valuation, led by Salesforce Ventures.
• November 2024 — Together AI announces a partnership with Hypertec to co-build a cluster of 36,000 Nvidia GB200 NVL72 GPUs (see below for what that hardware actually is).
• February 2025 — Together AI raises a $305 million Series B at a $3.3 billion valuation, led by General Catalyst and co-led by Prosperity7 Ventures. Annualized revenue is reported at just over $100 million, up from around $30 million a year earlier.
• June 2025 — Together AI announces a partnership with Hypertec and 5C Group to deploy up to 100,000 Nvidia GPUs across Europe through 2028.
• September 2025 — Together Instant Clusters reaches general availability, automating GPU cluster provisioning from single nodes to hundreds of interconnected GPUs. The company also goes live with Sweden-based infrastructure to serve European customers.
• Early 2026 — Reports emerge of Together AI seeking roughly $1 billion in new funding at a $7.5 billion valuation, with annualized revenue reportedly approaching $1 billion.
• July 1, 2026 — Together AI closes an $800 million Series C at an $8.3 billion valuation, led by Aramco Ventures, with annual bookings reported above $1.15 billion.

Products and Platform

Together Inference is the company's flagship API, giving developers access to more than 200 open-source models with what the company claims is 2 to 3 times faster performance than comparable hyperscaler solutions. Under the hood, it relies on proprietary optimizations built on research like FlashAttention and the newer FlashAttention 4, along with tools the company calls the Together Kernel Collection.

GPU Cloud and Instant Clusters give developers self-service access to Nvidia hardware, scaling from a single 8-GPU node up to clusters with hundreds or even tens of thousands of interconnected GPUs, using both Hopper-generation and newer Blackwell chips.

Fine-Tuning tools let teams adapt open models like Llama or DeepSeek to their own data without building a training pipeline from scratch.

Dedicated Clusters, marketed for token-heavy or latency-sensitive workloads, give enterprise customers guaranteed, isolated capacity rather than shared infrastructure.

Enterprise AI Platform wraps all of the above in the security, compliance, and support layer that large organizations require, including opt-out privacy controls for sensitive workloads.

Customers tend to cite three consistent reasons for choosing Together AI over larger cloud providers: significantly lower inference costs, faster response times for real-time applications, and the flexibility to swap between open models without being locked into a single vendor's ecosystem.

Customers

Together AI has disclosed a range of customers across its funding announcements and case studies, including Salesforce, Zoom, SK Telecom, Cognition, Zomato, Krea, Cartesia, Hedra, and The Washington Post.

Zomato, the Indian food delivery giant, built an AI customer support system on Together AI's platform that reportedly doubled customer satisfaction scores while scaling to more than 1,000 messages per minute, according to Together AI's own customer case studies.

Cartesia, a voice AI company, uses the platform to power ultra-low latency speech models. Pika Labs and Dippy AI have used dedicated endpoints for AI video and companion applications, with Dippy reportedly scaling to more than 4 million tokens processed per minute.

Decagon, another customer cited in reporting from Tech Funding News, reportedly cut its inference costs by roughly six times after moving workloads onto Together AI's platform.

The common thread across these customers is that none of them wanted to build and maintain their own GPU infrastructure. They wanted the performance of a well-optimized AI stack without the overhead of running it themselves.

Growth Strategy

Together AI's growth strategy rests on a handful of deliberate bets, each reinforcing the others.

Open-source first. Rather than competing to build the smartest proprietary model, the company positioned itself as the best place to run whichever open model wins. That's a hedge against picking the wrong horse in a fast-moving model race.

Developer-first distribution. By making it simple, and often free at small scale, to call an API and get a working model response, Together AI built a bottom-up user base of individual developers long before it started closing enterprise contracts. The company says its platform serves over a million developers.

Aggressive pricing on inference. Together AI has claimed that customers can cut inference costs anywhere from six to sixty times compared to closed-model rivals, according to figures reported by Tech Funding News. Even taking the lower end of that range seriously, it's a meaningful cost advantage for any company running AI at real scale.

Performance engineering as a moat. Having Tri Dao, the creator of FlashAttention, as chief scientist gives the company genuine technical credibility. Innovations like ThunderAgent and ATLAS-2 are pitched as delivering multiples of throughput improvement for demanding workloads.

Capacity expansion tied to demand. Rather than overbuilding speculative capacity, the company has tied major infrastructure commitments, like the 36,000-GPU Hypertec cluster or its European expansion with 5C Group, to visible customer demand and geographic gaps, particularly around data residency requirements in Europe.

Why Investors Are Betting on Together AI

The investment thesis behind Together AI isn't complicated, even if the technology underneath it is.

AI inference, the process of actually running a trained model to generate an answer, is projected to become a far bigger market than training ever was, simply because every single AI interaction from every user requires it. Training happens once. Inference happens billions of times a day.

Enterprise adoption of AI is accelerating, and a growing number of large organizations don't want to depend entirely on one or two closed-model vendors for something as central as their AI strategy. According to McKinsey research cited in coverage from Tech Funding News, nearly three in four organizations expect to increase their use of open-source AI over the coming years.

Open-source models like Meta's Llama and DeepSeek's R1 have closed much of the performance gap with proprietary systems, while offering more control, transparency, and often dramatically lower cost. That shift benefits whoever provides the best infrastructure for running those models, regardless of which specific model ends up on top.

There's also a capital efficiency argument. Together AI hasn't spent billions building its own foundation model from scratch, the way OpenAI or Anthropic have. Its capital goes almost entirely into infrastructure and optimization, a more asset-backed, less speculative kind of spending.

Finally, there's the broader macro backdrop. AI captured close to half of all global venture funding in 2025, according to data from Crunchbase cited by Tech Funding News, and infrastructure "neocloud" providers have become one of the hottest subcategories within that boom. Together AI's rapid valuation growth, from $1.25 billion to $8.3 billion in roughly two years, reflects investors racing to back the picks-and-shovels layer of the AI economy before it consolidates.

Competition

Together AI operates in a crowded field, but its specific positioning, open-source focus plus deep performance engineering, sets it apart from most rivals.

Together AI's real differentiator is focus. It has never built a proprietary foundation model, and it isn't trying to. That singular focus on being the best place to run other companies' open models, rather than splitting attention between chips, proprietary models, and infrastructure, is what several outlets, including Tech Funding News, have pointed to as its clearest strategic advantage over more diversified rivals.

Challenges

None of this comes without real risk.

GPU availability remains a persistent constraint. Together AI, like every player in this space, depends on Nvidia's production and allocation decisions. A supply shock or a shift in Nvidia's priorities could squeeze margins or growth plans quickly.

Capital intensity is brutal. Building and running GPU data centers costs enormous amounts of money, and much of that spending happens before revenue catches up. The company's push to scale compute infrastructure by roughly 50 times over the next five years, as reported following the Series C, will require sustained access to capital, not just this one round.

Competition is intensifying, not easing. Every major cloud provider, plus a growing list of well-funded neoclouds, is chasing the same enterprise customers. That kind of competition tends to compress pricing over time, which cuts directly against a business built partly on being the cheaper option.

The pace of AI innovation is its own risk. Model architectures, hardware generations, and optimization techniques are all shifting fast. A company that's the fastest at running today's leading open models has to keep reinventing that advantage every few months, not every few years.

Enterprise expectations for uptime, security, and compliance rise every year, and meeting those expectations at massive scale is a genuinely different operational challenge than serving individual developers.

Financial Growth

Together AI's revenue figures come from a mix of company statements and media reporting, and it's worth being precise about which is which.

According to Bloomberg reporting cited by Nasdaq, Together AI's annualized revenue was around $30 million in February 2024, and had grown to just over $100 million by February 2025.

Sacra, a private company research firm, has reported 2024 revenue of approximately $130 million, which would put the company at roughly a 9.6 times revenue multiple at its $1.25 billion valuation from that period. This is a reported estimate, not a confirmed company disclosure.

By early-to-mid 2026, reports citing The Information described annualized revenue approaching $1 billion, more than three times its level from mid-2025. Around the same period, Together AI's own disclosures, tied to its Series C announcement, cited annual bookings of over $1.15 billion.

It's worth being precise about the distinction. Bookings reflect contracted future spend, not necessarily revenue already recognized on the company's books. Together AI has not published audited financial statements, and as a private company, it isn't required to. Readers should treat any specific revenue or profitability figure not directly confirmed by the company as a reported estimate rather than a verified fact.

Together AI has not disclosed whether it is currently profitable, and no public source in this research confirms a profitability figure either way.

Future Outlook

The next few years will likely test whether Together AI's bet on open-source infrastructure was a temporary edge or a durable position.

AI inference is widely expected to keep growing faster than training spend, simply because every deployed AI application generates ongoing inference demand. If that trend holds, companies positioned specifically around inference efficiency stand to benefit disproportionately.

Enterprise AI adoption is still early. Most large organizations are only beginning to move AI workloads from pilot projects into full production, and that transition typically comes with much larger, longer-term infrastructure contracts.

Sovereign AI is an emerging theme worth watching. Governments and regions increasingly want AI infrastructure and data processing to happen within their own borders, for both security and regulatory reasons. Together AI's European expansion, including its Sweden infrastructure and its partnership with Hypertec and 5C Group to deploy up to 100,000 GPUs across the continent, positions it to capture some of that demand.

Global expansion more broadly will matter. AI demand is not confined to North America, and infrastructure providers that can offer low latency and data residency compliance across multiple regions have a real edge with international enterprise customers.

Open-source AI itself continues to evolve quickly, with new model releases from labs like Meta and DeepSeek arriving at a rapid pace. Together AI's fortunes are tied closely to the continued strength and competitiveness of the open-source ecosystem as a whole.

Agent infrastructure, the tooling needed to support AI systems that take multi-step actions rather than just answering single questions, is an emerging category the company has already started addressing through products like ThunderAgent, aimed at agentic workloads.

The risks are real too. Heavier competition, tighter GPU supply, and the sheer capital intensity of this business could all pressure margins even as revenue grows. But the underlying demand signal, enterprises wanting cheaper, faster, more flexible AI infrastructure, looks durable rather than temporary.

Entrepreneurial Lessons

Together AI's rise offers concrete lessons for founders building in infrastructure, deep tech, or any capital-intensive category.

1. Pick a side of the market others are ignoring. While most attention went to proprietary foundation models, Together AI bet on the unglamorous but massive opportunity of making open models actually usable at scale.

2. Founder technical depth is a fundraising asset. Having a Stanford professor, an ETH Zurich researcher, and the creator of FlashAttention as co-founders gave investors real confidence long before there was meaningful revenue to point to.

3. Timing matters, but so does surviving until timing works in your favor. Together AI launched months before ChatGPT, and its early bets on efficient inference happened to land right as the whole industry started caring about inference costs.

4. Don't try to do everything. Together AI never built its own foundation model. That focus let it become excellent at one specific layer of the stack instead of mediocre across several.

5. Let performance benchmarks do your marketing. Claims like "2 to 3 times faster inference" or "60 times lower cost" are more persuasive to technical buyers than any slogan could be.

6. Raise in stages that match your proof points. Each round, from the $20 million seed to the $800 million Series C, arrived alongside a clearer, bigger proof point: more developers, more revenue, more enterprise logos.

7. Strategic investors can be as valuable as the capital itself. Bringing in Nvidia, Salesforce Ventures, and Aramco Ventures wasn't just about money. Each of those investors opens doors to hardware access, enterprise customers, or global expansion.

8. Capital-intensive businesses need patient, aligned investors. GPU infrastructure isn't a business you can bootstrap. Together AI's willingness to raise large rounds repeatedly reflects an honest acceptance of what this category actually requires.

9. Build credibility with research, not just marketing. Publishing genuine technical work, like FlashAttention and the Together Kernel Collection, gave the company authority with the developers it needed to win over first.

10. Expand geographically only when demand justifies it. Together AI's move into Europe followed clear signals, data residency requirements and specific customer demand, rather than expansion for its own sake.

Key Takeaways

• Together AI was founded in June 2022 by Vipul Ved Prakash, Ce Zhang, Chris Ré, Percy Liang, and Tri Dao, betting that open-source AI infrastructure, not proprietary models, was the biggest opportunity in the AI boom.
• The company has never built its own foundation model. Its business is the cloud infrastructure that lets others run open-source models efficiently.
• Together AI's valuation climbed from $1.25 billion in March 2024 to $3.3 billion in February 2025 to $8.3 billion in July 2026, following an $800 million Series C led by Aramco Ventures.
• The company reports annual bookings above $1.15 billion, though this figure reflects contracted future spend, not confirmed audited revenue.
• Key products include Together Inference, GPU Cloud and Instant Clusters, fine-tuning tools, dedicated clusters, and an enterprise AI platform.
• Competitors include CoreWeave, Lambda, Fireworks AI, Nebius, Crusoe, and, in a different sense, closed-model labs like OpenAI and Anthropic.
• Risks include GPU supply constraints, intense competition, and the sheer capital intensity of building AI infrastructure at scale.

FAQ

What is Together AI? Together AI is a San Francisco-based AI infrastructure company, founded in 2022, that provides GPU cloud computing, inference, training, and fine-tuning services built specifically around open-source AI models.

Who founded Together AI? Together AI was founded by Vipul Ved Prakash (CEO), Ce Zhang (CTO), Chris Ré, Percy Liang, and Tri Dao (Chief Scientist).

How does Together AI make money? Together AI generates revenue through usage-based pricing for its GPU cloud and inference API, along with enterprise contracts for dedicated clusters, fine-tuning services, and its broader enterprise AI platform.

What is Together AI's valuation? As of its Series C round in July 2026, Together AI was valued at $8.3 billion, according to TechCrunch, up from $3.3 billion in February 2025.

Who invested in Together AI? Investors across Together AI's funding rounds include Lux Capital, Kleiner Perkins, Nvidia, Salesforce Ventures, General Catalyst, Prosperity7 Ventures, Coatue, Emergence Capital, Vista Equity Partners, and Aramco Ventures, which led the most recent Series C.

Is Together AI profitable? Together AI has not publicly disclosed profitability figures. As a private company, it is not required to release audited financial statements, and no confirmed profitability data is currently available.

What products does Together AI offer? Together AI's core products include Together Inference (its API for running open-source models), GPU Cloud and Instant Clusters, fine-tuning tools, dedicated clusters for enterprise workloads, and a broader enterprise AI platform.

Why is Together AI growing so quickly? Together AI's growth reflects rising enterprise demand for open-source AI as a lower-cost alternative to proprietary models, combined with the company's deep technical focus on inference speed and cost efficiency, and a broader surge in venture capital flowing into AI infrastructure.

Related reading:

OpenRouter Success Story

Skyroot Success Story

Sarvam AI Success Story