Baseten Raises $1.5B: Why AI Inference Is the New Gold Rush

AI inference startup Baseten is on the verge of raising a staggering $1.5 billion at a $13 billion valuation, according to a June 18 report from the Wall Street Journal covered by TechCrunch. This comes just five months after the company announced a $300 million Series E at a $5 billion valuation — representing a 160% increase in valuation in under six months.

The company’s rapid growth reflects what industry observers have started calling the “inference gold rush.” Inference is what happens after a user types a prompt: the AI model processes the input and generates a response. For video creators using AI tools, inference is the moment the AI generates a clip, applies an effect, or creates a transition. The faster and cheaper inference gets, the more powerful AI video tools become.

Baseten’s approach is notable because it routes AI requests to the best model for each task, often using competent open-source alternatives to reduce costs. This matters for video creators because video generation is one of the most compute-intensive AI workloads, and companies that optimize inference costs directly affect what creators pay for AI video tools. For more on AI video tools and platforms, see our AI video tools coverage.

What Is Inference and Why Is It the Gold Rush?

To understand why investors are pouring billions into inference companies, you need to understand the difference between training and inference. Training is when an AI model learns from data — it is expensive, one-time, and done by the companies that build the models. Inference is when a trained model responds to a user’s request — it happens constantly, every time anyone uses an AI tool, and the costs add up fast.

Video inference is particularly expensive. Generating a single ten-second AI video clip can require thousands of inference computations, each one processing visual information at a level of detail that text models never touch. As AI video tools become more popular, the demand for inference compute is exploding faster than the supply of available hardware can keep up.

This supply-demand imbalance is what VCs are betting on. Baseten is one of several companies building the infrastructure layer that routes inference requests to the most cost-effective model for each task. Instead of sending every request to an expensive frontier model, Baseten’s system evaluates the task and picks the right tool — often using open-source models that are 10-100x cheaper than the big proprietary models for specific tasks.

According to the Next Web’s analysis, the inference market is projected to grow from $15 billion in 2025 to over $60 billion by 2028, driven primarily by video and multimodal AI applications. That growth rate — roughly 40% annually — explains why investors are pricing Baseten at a $13 billion valuation so soon after its last round.

What the Split-Priced Round Tells Us

One detail worth noting: the Wall Street Journal reports that this is a split-priced round. Some investors are coming in at a $13 billion valuation, while others are at $11 billion. This is a tactic startups are using to boost their headline valuation while still offering better terms to strategic investors.

For video creators, the split pricing is a signal that the inference market is real but not yet stable. The gap between the two pricing tiers — roughly 15% — suggests that even the investors closest to the deal do not fully agree on what Baseten is worth. In a mature market, valuations converge. In a gold rush, they diverge because nobody knows exactly how big the opportunity is.

The round is said to be co-led by Spark Capital, Sands Capital, Altimeter Capital, and Wellington Management — a lineup of blue-chip investors that signals confidence in the inference thesis even if the exact valuation is still being debated.

How This Affects AI Video Tools

Baseten’s massive fundraise has direct implications for video creators and businesses that use AI video tools. The company’s technology routes requests to the best-for-task model, particularly competent open-source alternatives that are significantly cheaper than proprietary models. If Baseten succeeds in making inference dramatically cheaper for video generation, the effect will ripple through every AI video tool on the market.

Here is the specific mechanism. AI video tools like Runway, Pika, and Adobe’s Firefly pay for inference compute every time a user generates a video. Those costs determine the price of the tool. When inference gets cheaper, tool prices drop, or the tools can offer higher quality for the same price. Baseten’s $1.5 billion raise is a bet that inference costs will fall, volume will rise, and the companies that control the inference layer will capture significant value.

For creators, this means the next 12-18 months could bring a significant drop in AI video generation costs. When a company raises $1.5 billion to make something cheaper, it usually succeeds. The capital allows Baseten to buy more hardware, negotiate better pricing, and optimize its routing algorithms — all of which drive down the cost per inference request.

According to Gartner’s 2025 AI infrastructure report, organizations that adopt optimized inference architectures can reduce AI compute costs by up to 45% compared to single-vendor dependency. Baseten’s model — routing to the cheapest capable model for each task — is exactly the kind of optimization that enables those savings.

What Creators Should Do

If you use AI video tools professionally, Baseten’s raise is a signal to stay invested in AI video generation. The infrastructure is getting cheaper, faster, and more capable. The tools you use today will almost certainly be more capable and more affordable in 12 months as inference costs drop.

Second, pay attention to which AI video tools partner with inference optimization companies like Baseten. Tools that optimize their inference costs will have room to improve quality without raising prices. Tools that run on expensive, single-vendor inference infrastructure will be under pressure to raise prices or limit features.

Third, consider the strategic implications. When inference becomes cheap enough, AI video generation moves from a niche capability used by early adopters to a standard feature used by mainstream creators. The creators who build workflows around AI video now will have a significant advantage when the cost barrier drops and the market expands.

The inference gold rush is happening whether you participate or not. The only choice is whether you benefit from the falling costs or get caught off guard by them.

Baseten’s Growth Trajectory in Context

To appreciate how unusual Baseten’s fundraise is, consider the timeline. In 2024, the company raised a $150 million Series D. In early 2026, it raised a $300 million Series E at a $5 billion valuation. Now, just months later, it is raising $1.5 billion at a $13 billion valuation. That is $1.95 billion in total funding in under two years — a trajectory that would have been unthinkable for an infrastructure startup five years ago.

The reason Baseten can raise at this pace is that its revenue is growing even faster than its valuation. According to the Wall Street Journal report, Baseten’s revenue has been doubling every six months as more AI applications move from experimental to production. When your customers include some of the fastest-growing AI startups in the world, your revenue growth compounds fast.

For video creators, the revenue growth matters because it means Baseten is not just a fundraising story. It is a real business with real customers and real revenue. The inference infrastructure Baseten is building will survive and scale regardless of whether the broader fundraising environment tightens.

What This Means for the Broader AI Video Ecosystem

The inference gold rush has a multiplying effect on the AI video ecosystem. When companies like Baseten raise massive rounds, they buy more GPUs, hire more engineers, and build more infrastructure. That infrastructure enables new AI video capabilities that were previously too expensive or too slow to be practical.

Consider what becomes possible when inference costs drop by 45% — the optimization Gartner’s report cites. Real-time AI video generation becomes feasible for live streaming. AI-powered video editing at high resolution becomes affordable for small creator teams. Personalized video content at scale becomes a realistic marketing tactic rather than an expensive experiment.

Each of these capabilities changes the competitive landscape for video creators. The creator who can generate personalized AI video ads at scale has an advantage over the one who cannot. The editor who uses AI inference to automate rotoscoping and keyframe generation works faster than the one doing it manually. The production company that builds its workflow around cheap inference will outbid competitors who still rely on traditional post-production methods.

The $1.5 billion flowing into Baseten is not just a financial story. It is infrastructure being built that will change how video is created in the United States and globally over the next three to five years.

Frequently Asked Questions

What is AI inference?

Inference is the process where a trained AI model responds to a user’s input. For video generation, inference is the computational work required to generate each frame, apply effects, and produce the final output.

Why is Baseten raising so much money?

Baseten is raising $1.5 billion at a $13 billion valuation to build and scale AI inference infrastructure. The company routes AI requests to the most cost-effective model for each task, reducing inference costs for its customers.

What does the inference gold rush mean for video creators?

Falling inference costs mean AI video generation will become cheaper and more accessible. Creators who build workflows around AI video now will benefit when the cost barrier drops and the market expands in the next 12-18 months.

What is a split-priced funding round?

A split-priced round means different investors pay different prices for their equity. In Baseten’s case, some investors enter at a $13 billion valuation while others enter at $11 billion. This signals that even investors disagree on the exact value of the company.

How fast is the inference market growing?

The inference market is projected to grow from $15 billion in 2025 to over $60 billion by 2028, driven primarily by video and multimodal AI applications — roughly 40% annual growth.

More AI Resources for Your Business