NVIDIA-backed Together AI secures $800m for inference infrastructure growth

Together AI has raised $800 million in Series C funding at an $8.3 billion post-money valuation as demand grows for infrastructure capable of running open artificial intelligence models at commercial scale. Aramco Ventures led the financing, with participation from Vista Equity Partners, General Catalyst, Emergence Capital, NVIDIA Corporation, March Capital, Pegatron Corporation and S Ventures, the investment arm associated with SentinelOne, Inc. The San Francisco company said annual bookings crossed $1.15 billion during the latest quarter and plans to expand its infrastructure capacity roughly 50-fold over five years. The new capital will support computing capacity, inference products, training services, reinforcement-learning tools and the wider infrastructure required to serve artificial intelligence developers and enterprises. The strategic question is whether Together AI can convert rapid bookings growth into defensible margins while competing against hyperscale clouds, specialised inference providers and closed-model companies with considerably larger balance sheets.

The new valuation represents an increase of approximately 152% from the $3.3 billion valuation attached to Together AI’s $305 million Series B in February 2025. The company had previously raised $102.5 million in a Series A round in 2023, making the latest transaction a substantial acceleration in both funding scale and investor expectations.

Together AI has not disclosed revenue, gross margin, operating losses, cash burn or the precise amount of computing capacity already contracted. Those omissions matter because annual bookings measure committed customer demand rather than revenue already recognised or cash already collected.

Why does Together AI’s $800 million Series C matter to the wider artificial intelligence market?

The financing reflects a change in where investors expect value to accumulate across artificial intelligence. The first phase of the current investment cycle focused heavily on companies creating foundation models. Attention is now shifting toward the infrastructure required to train, customise and serve those models efficiently inside commercial applications.

Together AI is targeting companies that want advanced artificial intelligence without depending entirely on a closed provider such as OpenAI or Anthropic. Its platform supports open models including DeepSeek, Nemotron, MiniMax and Kimi, alongside infrastructure for model training, fine-tuning, reinforcement learning and production inference.

That positioning matters because model costs can consume a significant portion of an application company’s revenue. A coding assistant, customer-service platform or research agent may appear commercially attractive until millions of daily requests create an unexpectedly impressive cloud bill.

Together AI argues that optimised open models can provide comparable performance at materially lower cost. Customers can also retain greater control over the model, deployment environment, proprietary data and product roadmap.

The round therefore represents a bet on open ecosystems rather than one particular model family. Together AI can potentially benefit as developers move among different open models, provided those workloads continue running through its infrastructure.

However, the company still depends on expensive semiconductors, data centres, energy and networking capacity. Open models may reduce customer dependence on a model provider, but they do not make the underlying computers free.

Can $1.15 billion in annual bookings justify Together AI’s $8.3 billion valuation?

The reported annual bookings figure gives investors a useful measure of commercial demand, but it is not equivalent to annual recurring revenue. Bookings can include multiyear contracts, future computing commitments and arrangements that will be recognised as revenue over time.

At the headline level, the $8.3 billion post-money valuation is approximately 7.2 times the $1.15 billion annual bookings figure. That appears less extreme than valuations assigned to some earlier artificial intelligence companies, although the comparison becomes less reassuring when bookings are converted into actual revenue and gross profit.

Infrastructure businesses frequently carry substantial pass-through costs. Together AI may invoice a customer for computing consumption, but a large portion of the payment may ultimately cover chips, cloud infrastructure, electricity, data-centre operations and networking.

Gross margin therefore matters more than booking volume alone. Together AI must demonstrate that its inference optimisation, software platform and infrastructure management create enough value above the cost of the underlying computing capacity.

The company’s growth rate also influences the valuation. Open-model usage across the industry has reportedly tripled over 12 months, while Together AI now serves thousands of paying customers. A rapidly expanding market can support a high valuation when a company gains share while improving unit economics.

The risk is that investors are capitalising future revenue too early. The valuation has increased by more than 150% since the Series B, placing pressure on Together AI to deliver infrastructure expansion, customer growth and margin improvement without another equally dramatic increase in operating costs.

The eventual public-market test will be stricter. Institutional investors will want audited revenue, customer-retention data, infrastructure commitments, stock-based compensation, gross margin and free cash flow before deciding whether $8.3 billion represented a conservative entry point or an expensive enthusiasm tax.

Why is open-model inference becoming a major alternative to closed artificial intelligence systems?

Closed-model platforms provide convenience because customers can access sophisticated capabilities through a standard application programming interface. However, the customer has limited control over model architecture, pricing, availability, safety restrictions and future product decisions.

Open models provide a different trade-off. Enterprises can select a model, adapt it using proprietary data, control where it runs and optimise the system around a specific commercial task.

This flexibility can reduce costs when an application generates high volumes of relatively predictable requests. A company may not need the most expensive frontier model for every customer interaction when a smaller specialised model can complete the task successfully.

Data control is another advantage. Financial institutions, healthcare companies, governments and industrial groups may prefer systems that can operate within dedicated or privately controlled environments.

Together AI is trying to remove the technical difficulty associated with this independence. Managing open models requires infrastructure orchestration, performance optimisation, monitoring, fine-tuning and reliable access to graphics processors.

The company’s opportunity grows if enterprises conclude that model ownership and portability are strategically important. It weakens if closed providers reduce prices rapidly enough to make infrastructure control less valuable.

The market is unlikely to become entirely open or entirely closed. Many companies will use premium closed models for difficult tasks while deploying smaller open models for specialised, high-volume applications. Together AI needs to become the infrastructure platform supporting that mixed architecture.

How will Together AI use the new funding to expand computing capacity by 50 times?

Together AI expects its infrastructure footprint to grow roughly 50-fold over the next five years. Reaching that target will require far more than purchasing servers.

The company must secure access to advanced graphics processors, data-centre space, electricity, cooling, networking equipment and long-term infrastructure financing. It must then integrate those resources into a service capable of delivering consistent performance across several locations.

Together AI already operates an artificial intelligence factory in Maryland in partnership with infrastructure provider 5C using NVIDIA B200 processors. Another facility in Memphis is being developed with NVIDIA GB200 and GB300 systems, while additional locations are expected to follow.

Geographic expansion can improve latency, resilience and access to electricity. It also creates operational complexity because each data centre may have different energy costs, networking arrangements, hardware configurations and regulatory requirements.

The $800 million round gives Together AI more negotiating power when reserving computing capacity. Suppliers may prioritise customers capable of making larger and longer-term commitments during periods when the newest chips remain scarce.

Those commitments can become liabilities when demand changes. Capacity must be secured before customers always know exactly how much computing they will consume. Together AI could face weak utilisation if bookings convert more slowly than expected or if newer chips make existing systems less competitive.

Infrastructure growth must therefore be paced against contracted customer demand. Expanding 50-fold sounds appropriately dramatic for artificial intelligence, but the economically important number is the proportion of that capacity that customers pay to use.

Can Together AI defend its position against hyperscale cloud providers and specialist rivals?

Together AI competes with Amazon Web Services, Microsoft Azure, Google Cloud and Oracle Cloud Infrastructure, all of which have enormous capital budgets, global data-centre networks and existing enterprise relationships.

These providers can bundle computing with databases, storage, cybersecurity, analytics and productivity software. They can also offer discounted infrastructure as part of wider commercial agreements.

Together AI must compete through specialisation. Its platform is designed specifically for artificial intelligence training and inference, allowing the company to focus on model performance, cost optimisation and developer experience rather than offering every category of enterprise technology.

Specialised rivals create another challenge. Baseten, Fireworks AI, Lambda, CoreWeave and other infrastructure companies are targeting overlapping parts of the artificial intelligence workload market.

Baseten recently raised $1.5 billion at a $13 billion valuation, showing that investors are willing to fund several potential winners simultaneously. That does not mean the market will support all current valuations once pricing competition intensifies.

Together AI’s research capabilities provide an important differentiator. Work associated with FlashAttention, inference optimisation and open training datasets can improve platform performance while strengthening credibility with developers.

Research leadership must translate into commercial advantages. A technically elegant optimisation has limited strategic value when a competitor incorporates a similar method or a hardware generation changes the cost equation.

Customer relationships may become the stronger moat. Together AI counts companies including Cursor, Cognition and Decagon among its users. Infrastructure embedded inside a commercial product can be difficult to replace when migration risks service disruption.

However, large customers also have negotiating leverage. Successful artificial intelligence companies may build internal infrastructure, negotiate directly with cloud providers or distribute workloads across several suppliers to avoid dependence on Together AI.

Why are Saudi-linked capital and energy strategy important to the Together AI round?

Aramco Ventures led the Series C, strengthening the relationship between artificial intelligence infrastructure and energy capital. Its Prosperity7 programme had previously co-led Together AI’s Series B financing.

The strategic logic extends beyond financial returns. Artificial intelligence data centres require enormous quantities of reliable electricity, making energy availability central to the next phase of computing expansion.

Saudi Arabia and other Gulf economies are investing in data centres, cloud infrastructure and artificial intelligence as part of wider efforts to diversify beyond traditional hydrocarbons. Capital providers connected to the region can support infrastructure development, market access and long-term computing projects.

Together AI may gain opportunities to expand capacity outside the United States or serve enterprises and government-backed technology programmes across the Middle East.

This creates a potentially valuable combination of software expertise, capital and energy infrastructure. It also exposes the company to geopolitical, regulatory and data-sovereignty considerations as computing capacity becomes more strategically sensitive.

The round illustrates how artificial intelligence fundraising is changing. Large infrastructure companies increasingly need investors capable of supporting multibillion-dollar capital programmes rather than conventional venture funds designed mainly to finance software teams.

Together AI’s investor group reflects that transition. Aramco Ventures brings energy-linked capital, Vista Equity Partners adds enterprise-software experience, NVIDIA Corporation supplies the dominant computing architecture, and Pegatron Corporation contributes hardware manufacturing exposure.

What do NVIDIA and SentinelOne share prices signal about investor sentiment toward the round?

NVIDIA Corporation participated in the Series C and remains central to Together AI’s infrastructure through its graphics processors and networking systems. NVIDIA shares traded near $196 on July 1, giving the company a market capitalisation of approximately $4.8 trillion.

NVIDIA stock had declined roughly 1.6% over five trading days and about 8.9% over one month. The shares remained within a 52-week range of approximately $152.97 to $236.54.

The recent pullback suggests investors are becoming more selective after an extraordinary expansion in artificial intelligence infrastructure spending. The market is not rejecting the long-term demand thesis, but it is questioning how rapidly spending can continue and which companies will capture acceptable returns.

Together AI provides NVIDIA Corporation with another channel through which developers consume its hardware. An independent artificial intelligence cloud can broaden NVIDIA Corporation’s ecosystem beyond the largest hyperscale providers.

SentinelOne, Inc. is represented through S Ventures, which also participated in the funding. SentinelOne shares traded around $17.75 on July 1, after gaining roughly 10% over five sessions and approximately 11% over one month.

SentinelOne remained within a 52-week range of $11.81 to $21.40. The improving trajectory reflects stronger sentiment toward artificial intelligence-powered cybersecurity, although the company’s valuation remains below earlier market peaks.

The investments are not individually large enough to transform either listed company’s financial profile. Their relevance lies in strategic positioning, as public technology companies seek exposure to the infrastructure providers enabling the next generation of artificial intelligence applications.

What risks could prevent Together AI from turning its funding into durable market leadership?

Capital intensity is the most immediate risk. Together AI needs to commit money to infrastructure before the full revenue generated by that infrastructure becomes visible.

Hardware obsolescence can weaken returns. A cluster built with current-generation processors may remain operational for years, but newer chips could offer better performance per unit of electricity and reduce the commercial value of older systems.

Pricing pressure represents another threat. Closed-model providers are reducing prices, open models are becoming more efficient and cloud companies can use infrastructure discounts to protect strategic customers.

Customer concentration may emerge as Together AI scales. A relatively small group of rapidly growing artificial intelligence companies could account for a meaningful share of bookings, exposing the platform to volatility if one customer changes suppliers or reduces spending.

Bookings conversion also requires scrutiny. Large commitments create confidence, but customers may negotiate ramp schedules, usage flexibility or termination provisions that delay revenue.

Regulation could affect the availability of chips, the movement of models and the countries in which Together AI can deploy computing capacity. Export controls and national-security rules are becoming part of ordinary infrastructure planning.

The company must also balance product breadth with focus. Training, inference, fine-tuning, evaluations and reinforcement learning create a comprehensive platform, but each category requires engineering and support.

Together AI has enough capital to expand aggressively. The harder task is ensuring that growth produces an independent, profitable company rather than a well-funded customer for chipmakers and data-centre operators.

What should investors watch after Together AI completes the $800 million financing?

Revenue disclosure will be the first major milestone. Investors need to understand how the $1.15 billion bookings figure converts into recognised revenue and cash receipts.

Gross margin will reveal whether Together AI’s optimisation technology creates meaningful economic value after paying for computing. Rising bookings accompanied by weak margins would suggest the company is purchasing growth through expensive infrastructure.

Utilisation rates should also be monitored. A larger capacity footprint creates value only when workloads generate sufficient revenue to cover equipment, energy and financing costs.

Customer diversification will indicate whether Together AI is becoming enterprise infrastructure or remains concentrated among artificial intelligence-native startups. Traditional companies may offer larger and more durable contracts but require longer sales cycles and stronger compliance capabilities.

The rollout of additional artificial intelligence factories will show whether the planned 50-fold capacity expansion is progressing. Project delays, power constraints or hardware shortages could limit growth despite strong customer demand.

A future financing round or initial public offering would provide another valuation test. Together AI’s $8.3 billion price already assumes that open models will capture a meaningful share of enterprise artificial intelligence spending.

The Series C gives management the resources to pursue that opportunity. It also removes the excuse that inadequate funding prevented execution.

Key takeaways on what Together AI’s $800 million funding means for open artificial intelligence

• Together AI has raised $800 million in Series C financing at an $8.3 billion post-money valuation.
• Aramco Ventures led the round, while Vista Equity Partners, NVIDIA Corporation, General Catalyst and other investors participated.
• The valuation has increased approximately 152% from the $3.3 billion level established during Together AI’s 2025 Series B.
• Annual bookings crossed $1.15 billion, although Together AI has not disclosed recognised revenue, gross margin or cash flow.
• The company plans to increase its infrastructure capacity roughly 50-fold over five years.
• Open models could help customers reduce inference costs and maintain greater control over proprietary data and model deployment.
• Together AI must compete with hyperscale cloud providers and heavily funded specialists including Baseten, CoreWeave and Fireworks AI.
• Computing commitments create utilisation, hardware-obsolescence and pricing risks despite the strong demand outlook.
• NVIDIA Corporation’s participation strengthens Together AI’s hardware ecosystem but does not remove dependence on advanced chip supply.
• The long-term valuation case depends on converting bookings into high-margin revenue rather than simply reselling expensive computing capacity.