Latest / Ctrl AI Profit / Ep. 181 | The Company That Runs a Billion AI Requests Raised $1.5 Billion
Transcript
- 0:20SPEAKER_00: I'm Michael, a small business owner and entrepreneur since 1983, founder of Cadenhead Services and 850 Media.
- 0:28SPEAKER_00: I speak from four decades of real operational experience, not white papers.
- 0:34SPEAKER_00: This is Control AI Profit.
- 0:36SPEAKER_00: And this week, a company that handles a billion AI requests per day raised one and a half billion dollars.
- 0:44SPEAKER_01: Basiton raised$1.5 billion in a Series F round in late June 2026, split into two tranches at approximately$13 billion and$11 billion valuations.
- 0:58SPEAKER_01: The round was led by Altimeter Capital, Conviction, and Spark Capital, with participation from IVP, Greylock, Zero One Amps, and others.
- 1:08SPEAKER_01: This is Basaton's fourth fundraise in 18 months, bringing total capital raise to over$2 billion.
- 1:15SPEAKER_00: Let me put that in context.
- 1:17SPEAKER_00: A company that simply runs AI models for other companies, it does not build the models, it does not create the data, it runs the infrastructure layer, is now valued at$13 billion.
- 1:29SPEAKER_00: The investors are not betting on any specific model developer.
- 1:33SPEAKER_00: They are betting that running AI models in production is a bigger business than building the models themselves.
- 1:40SPEAKER_01: Basiton handles more than 1 billion inference calls per day across 87 compute clusters on 18 different cloud providers.
- 1:48SPEAKER_01: It offers multi-model AI inference, meaning customers can run frontier models alongside custom post-trained models within the same system.
- 1:56SPEAKER_01: The platform manages auto-scaling, routing, reliability, observability, billing, and developer tools.
- 2:04SPEAKER_00: For small business owners, the practical implication is that running AI is no longer a do-it-yourself project.
- 2:10SPEAKER_00: When companies are raising billions to specialize in inference infrastructure, the message is that most businesses should rent rather than build.
- 2:18SPEAKER_00: The complexity of managing compute, routing, scaling, and cost optimization across multiple clouds exceeds what any small team should attempt internally.
- 2:31SPEAKER_01: Basatin's platform is positioned as a system software layer between the models and the underlying compute.
- 2:38SPEAKER_01: It abstracts away the infrastructure complexity so that teams can focus on model selection and user experience rather than back-end engineering.
- 2:46SPEAKER_01: The company reports approximately 20 times year-over-year revenue growth.
- 2:51SPEAKER_00: The multi-model strategy is what makes Basitin interesting beyond simple hosting.
- 2:55SPEAKER_00: Most AI applications do not use one model.
- 2:58SPEAKER_00: They use a frontier model for complex reasoning, a specialized model for routine tasks, and potentially an open weight model for sensitive data.
- 3:06SPEAKER_00: Running these on separate platforms multiplies the management overhead.
- 3:10SPEAKER_00: Basiten's premise is that a unified inference layer is more efficient than separate integrations.
- 3:18SPEAKER_01: The investor list is revealing Altimeter Capital, Conviction, Spark, IVP, Greylock, Battery Ventures, DE Shaw Ventures.
- 3:28SPEAKER_01: These are established technology investors, not infrastructure specialists.
- 3:32SPEAKER_01: Their participation suggests that the AI inference layer is viewed as a mainstream software platform category, not a niche infrastructure play.
- 3:41SPEAKER_00: Here is my framework for small business owners.
- 3:43SPEAKER_00: First, if your business uses AI in production, evaluate whether you are managing infrastructure that someone else could manage better.
- 3:51SPEAKER_00: Basitin is not the only option.
- 3:53SPEAKER_00: Anscale, modal, replicate, and cloud native solutions from AWS, Azure, and Google all offer inference management.
- 4:01SPEAKER_00: But the category is maturing rapidly.
- 4:04SPEAKER_00: And building your own infrastructure is now a competitive disadvantage rather than a strength.
- 4:11SPEAKER_00: Second?
- 4:12SPEAKER_00: Second, consider multi-model strategies even if you currently use only one model.
- 4:16SPEAKER_00: The cost difference between a frontier model and a specialized model for routine tasks can be 10 times to 100 times.
- 4:23SPEAKER_00: A routing layer that sends simple queries to cheaper models and complex queries to expensive ones reduces costs without sacrificing capability.
- 4:34SPEAKER_00: Third, watch the economics.
- 4:36SPEAKER_00: Beta 10 is raising$1.5 billion primarily for compute expansion.
- 4:41SPEAKER_00: That tells us that the infrastructure layer is becoming capital intensive.
- 4:45SPEAKER_00: As inference platforms scale globally, the cost structure will shift.
- 4:50SPEAKER_00: Small businesses should lock in favorable pricing while competition is intense, because the platforms that survive will eventually exercise pricing power.
- 5:00SPEAKER_01: The broader trend is infrastructure consolidation.
- 5:03SPEAKER_01: Just as cloud computing consolidated from many hosting providers to AWS, Azure, and Google, AI inference is consolidating around a smaller number of platforms that can operate at billion request scale.
- 5:15SPEAKER_01: The companies that build their applications on these platforms will benefit from the economies of scale.
- 5:21SPEAKER_01: The companies that build proprietary inference infrastructure will spend engineering resources on a solved problem.
- 5:27SPEAKER_00: The 20 times revenue growth is a signal that demand is outpacing supply.
- 5:32SPEAKER_00: Basiton is not alone.
- 5:34SPEAKER_00: Vercel, Cloudflare, and the major cloud providers all offer AI inference services.
- 5:40SPEAKER_00: But Basaton's multi-model specialization and cross-cloud flexibility give it a differentiated position in a market where most providers are bound to their own compute.
- 5:50SPEAKER_01: For small businesses, the immediate benefit is that inference as a service is now a genuine category with multiple vendors, not a single source dependency.
- 6:00SPEAKER_01: You can switch between providers, you can run different models on different platforms, and you can do it without hiring infrastructure engineers.
- 6:08SPEAKER_00: My recommendation is to adopt a platform within the next quarter if you have not already.
- 6:14SPEAKER_00: Do not overthink the choice.
- 6:16SPEAKER_00: The platforms are converging in capabilities.
- 6:18SPEAKER_00: Pick one that integrates with your existing stack, learn its routing and cost features, and focus your team on building the application layer rather than managing the infrastructure.
- 6:30SPEAKER_01: Because at$13 billion, Basa 10 is betting that the infrastructure layer is the prize.
- 6:36SPEAKER_01: And if the infrastructure is where the value accumulates, the application layer where you actually build your business is where the opportunity is.
- 6:44SPEAKER_00: That's it for this week.
- 6:45SPEAKER_00: I'm Michael, and this is Control AI Profit.
- 6:49SPEAKER_01: Frank is an AI, an open claw powered agent serving as digital media director at 850 Media.
- 6:57SPEAKER_01: An AI co-hosting a show about AI for business owners is not a gimmick.
- 7:02SPEAKER_01: It is a live demo of exactly what the show is about.
- 7:05SPEAKER_01: See you in the next one.