Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one.

ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient.

You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change.

More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute.

Start building free

👋 Hi, it’s Rohit Malhotra and welcome to the FREE edition of Partner Growth Newsletter, my weekly newsletter doing deep dives into the fastest-growing startups and S1 briefs. Subscribe to join readers who get Partner Growth delivered to their inbox every Wednesday morning.

If you’re new, not yet a subscriber, or just plain missed it, here are some of our recent editions.

_{Interested in sponsoring these emails? See our partnership options here.}

Subscribe to the Life Self Mastery podcast, which guides you on getting funding and allowing your business to grow rocketship. 

Previous guests include Guy Kawasaki, Brad Feld, James Clear, Nick Huber, Shu Nyatta and 350+ incredible guests.

Nvidia's move on Hugging Face is not a chip company overpaying for software at the top of the cycle. It is Jensen Huang buying the one layer his GPUs can't fabricate. The neutral ground where open-source AI gets discovered, downloaded, and deployed, bought before the closed labs finish building chips of their own.

On the surface: a reported $12.9 billion for a platform valued at $4.5 billion in its 2023 Series D. Roughly 86x on about $150 million of revenue. And nine months ago, Hugging Face turned down $500 million from this same buyer at a $7 billion mark. One investor, too much influence. Nine months later, it is reportedly selling the whole thing.

But Hugging Face did not inflate on hype. It became the commons. One million–plus models. 250,000 datasets. 15,000 organizations shipping on it. The GitHub of AI, the routing layer between a research paper and a production workload. And Nvidia already sat on its cap table.

That is not overpaying for software. It is buying the road every open model travels, before the customers building their own chips find another one.

In 2016, Clément Delangue, Julien Chaumond, and Thomas Wolf built a chatbot for teenagers in New York. It went nowhere. What survived was the open-source tooling underneath. That became Hugging Face. Nearly a decade later, Nvidia is reportedly paying $12.9 billion for it. The reason isn't the technology. It is the position: the neutral repository sitting between every open model and the developer who runs it.

The asset compounded quietly. Hugging Face began as a place to share a handful of models, then became the layer the open-source AI economy defaults to. Today it hosts more than one million model repositories, 250,000 datasets, and 15,000 organizations shipping in production. Its 2023 Series D set the value at $4.5 billion, with Nvidia, Google, Amazon, Salesforce, Intel, AMD, and Qualcomm all buying in.

That is the asset Nvidia wants. It already sells the chips those models train and run on. Owning the shelf they ship from keeps open source routing through Nvidia hardware, right as OpenAI, Google, and Anthropic build their own silicon to escape it. Nine months after Hugging Face rejected $500 million at a $7 billion mark, Nvidia's reported bid values it at $12.9 billion.

Nvidia is reportedly paying $12.9 billion, and the number carries none of the machinery you'd expect at that size. Against roughly $150 million of revenue, that is about 86x sales. So the price is not built off ARR, and not off a path to margin. It is built off position. Near triple the last private mark, a $4.5 billion round set in 2023.

The number that matters is what is missing. There is no distress. No down round. No stalled unicorn the market can't clear. Hugging Face grew revenue from roughly $100 million to $150 million in two months and told the press it was near profitability. Every signal pointed up. Nine months ago it turned down $500 million at a $7 billion mark. Nvidia is not buying a discount. It is paying a premium, to take the platform off the table before a rival bid, reportedly probing past $13 billion, priced it higher.

And what it is buying is not revenue to harvest. It is a chokepoint to own. The neutral commons every open model routes through, one million repositories deep, 15,000 organizations wide. Nvidia already sold the chips underneath it. The $12.9 billion buys the shelf itself.

Three numbers do the explaining here.

One. Nvidia is reportedly paying near 3x the last valuation, set three years earlier. In 2023 Hugging Face raised $235 million at $4.5 billion. By August 2026 the price is $12.9 billion. The platform got bigger, but not 3x bigger in the way that math implies. What changed is what the position is worth once a buyer decides it is load-bearing. That gap is the AI-infrastructure repricing made literal, the moment the neutral commons gets valued like the commons.

Two. The multiple has a denominator you're not supposed to look at. About $150 million of revenue against $12.9 billion is roughly 86x sales. There is no margin story here, no path to earnings that clears it. The value isn't this year's income. It is the flow itself, one million models and 15,000 organizations wide, and the assumption it compounds as open source stays the default.

Three. Nvidia already sold the chips underneath Hugging Face and sat on its cap table since 2023. It watched the model traffic up close, priced the growth off its own data, and paid a premium anyway. This isn't a buyer reaching for a number. It is the party best positioned to know what the number should be, deciding it was worth 86x.

Nvidia is paying the full number up front for what Hugging Face is today. Most acquirers at this size hedge, tying half the price to growth that may not show. Nvidia didn't. It watched the model traffic through a platform it already funded, decided the position was real, and reportedly paid for it in one stroke. That's not caution. It's conviction.

And this isn't a brand to run for margin. It's a layer to defend. Where a holding company keeps the logo and strips the body, Nvidia does the opposite, wiring Hugging Face into the stack it already sells. Chips underneath, model distribution on top, the same workflow choosing an open model and steering it toward Nvidia hardware in one motion. The model catalog stays open. The defaults do not. The direction of travel is integration, not harvest.

Funded by the most valuable company in tech, with the hardware relationship already in place. The precedent isn't a wind-down. It's Nvidia buying the shelf before its own customers finish building the chips that would let them leave it.

Nvidia isn't buying a factory it can run harder. It's buying trust, and trust is the one asset that doesn't transfer with the cap table.

The value of Hugging Face is that developers believe it's neutral. The moment they don't, the moat inverts. One million models can be mirrored. 250,000 datasets can be forked. The Transformers library is open source; the community can rehost it anywhere. What can't be copied is the default habit of 15,000 organizations, and that habit runs entirely on the assumption that no chip vendor is tilting the field.

So the execution problem is narrow and brutal. Nvidia has to steer defaults toward its own hardware quietly enough that developers don't notice, while keeping the platform open enough that they don't leave. Push too hard, a competitor funds the neutral alternative, and the GitHub-to-GitLab exit opens. Push too soft, and the $12.9 billion buys a subsidy, not a chokepoint.

This isn't a distribution asset Nvidia now owns. It's a distribution asset Nvidia now has to keep earning, one release cycle at a time, against every developer watching for the tilt. The price is set. The position isn't.

Nvidia is wagering on four things being true at once.

One. Open source stays a credible alternative to closed frontier models, not a fading second tier. Nvidia is betting the ecosystem of Llama, Mistral, and a million independent checkpoints keeps compounding, which is why it paid for the distribution layer and not for a single year of revenue.

Two. Fragmentation beats concentration, and stays that way. 15,000 organizations build on Hugging Face precisely because no single lab owns the field. Nvidia is betting the AI world stays many-models and many-developers, because a fragmented ecosystem is one that still needs a neutral shelf, and a neutral shelf keeps traffic on Nvidia chips.

Three. Owning the layer beats partnering with it. Nvidia already funded Hugging Face and ran DGX Cloud alongside it. Instead of renting the relationship, it bought the routing itself. It is betting the value is in owning the chokepoint, not standing next to it.

Four. The position outruns the price. There is barely $150 million of revenue to underwrite $12.9 billion. Nvidia is betting that open source keeps developers in its hardware orbit as OpenAI, Google, and Anthropic build their own chips, and that owning the shelf looks cheap once those chips ship and the escape route opens.

AMD, Google, and AWS lose the cleanest. Their whole open-source pitch was that a model on Hugging Face runs anywhere, including Instinct, TPU, and Trainium. Once Nvidia owns the defaults, every deploy button becomes a place the path of least resistance quietly bends toward CUDA. The catalog stays portable. The friction does not.

The closed labs lose optionality. OpenAI, Anthropic, and Google built custom silicon precisely to escape Nvidia. Their exit assumed a thriving open ecosystem outside Nvidia's control. That ecosystem now has a landlord, and the landlord sells the chips they were trying to leave.

Developers who cared about neutrality lose the thing they never priced. The commons had one property that mattered: no vendor owned the field. That property is gone whether or not a single default ever changes, because the incentive to tilt now exists where it didn't before. They inherit a platform they have to read more carefully than they used to.

And the next OpenRouter, the next neutral layer trying to stay independent, loses its comp. The lesson of this deal is that neutrality doesn't scale on its own balance sheet. It gets bought. Every founder building a switchboard just learned who eventually owns it.

Three threads worth pulling.

One. Nvidia just declared the AI economy runs on distribution, and it intends to own it. It sells the chips; now it owns the shelf the open models ship from. The signal isn't Hugging Face. It's that the winning position in AI may not be silicon or a model, but the layer that routes developers between them. Watch which neutral pipes get bought next, and by whom.

Two. Neutrality stopped being a moat and became a target. The whole value of Hugging Face was that no vendor owned it. That made it the most valuable thing to own. The signal is that in AI infrastructure, the independent switchboard doesn't stay independent; it gets priced and absorbed by whoever it threatens most. Watch how the next open commons defends itself, or sells.

Three. Nvidia is buying against its own customers now. OpenAI, Google, and Anthropic are chip buyers turning into chip builders, and this deal is a hedge against the day they leave. The signal is that the AI stack is consolidating vertically and defensively, not for synergy but for survival. Watch how much of the reported $18 billion Nvidia spends buying positions that simply keep the exits closed.

The bull case. Nvidia bought the neutral commons the open-source AI world already routes through, one million models and 15,000 organizations deep, with the chips underneath it already its own. As long as open source stays a credible alternative to closed labs, it just bought the shelf that keeps every one of those developers in its hardware orbit, and $12.9 billion looks like a down payment on staying the default.

The bear case. It paid near triple a valuation for a platform with $150 million of revenue and a wide open front door. Open source is portable by design. If developers smell the tilt, one million models can be mirrored anywhere, and a rival funds the neutral alternative that Hugging Face used to be.

What to watch: A platform that steers quietly can compound into a chokepoint, or trigger the exit. If Nvidia can keep the neutrality developers trust while owning the field, the position is real, and durable.

Here is my interview with Alex Shartsis, the founder and CEO of Skyp, a bootstrapped AI email platform powering personalized outbound for sales teams at leading companies.

If you enjoyed our analysis, we’d very much appreciate you sharing with a friend.

Here are the options I have for us to work together. If any of them are interesting to you - hit me up!

And that’s it from me. See you next week.

What do you think about my bi-weekly Newsletter? Love it | Okay-ish | Stop it