Potentially AI has pulled off a neat trick for a young company, persuading two of the largest names in artificial intelligence to work with it at once.
The AIM-listed group has joined the BytePlus Partner Network, the enterprise programme run by ByteDance's BytePlus arm, days after winning a place in the AI tier of the Google for Startups Cloud Program.
For a business that describes itself as a platform built to work across every model provider, landing Google and ByteDance is a genuine validation of the pitch.
The trouble is that the announcement tells investors a great deal about Potentially's suppliers and almost nothing about its customers.
Why neutrality is the strategy
The logic behind the deals is worth understanding, because it is the heart of the company's proposition.
Potentially's positioning is "one platform, all models", meaning it does not build its own artificial intelligence but instead knits together the systems made by others.
That model-agnostic stance is precisely what lets it sign up rivals without friction.
A company wedded to a single provider could not credibly court Google and ByteDance simultaneously, but a neutral aggregator has every incentive to befriend all of them, and they have an incentive to be inside its shop window.
Chief operating officer Oliver Yonchev framed the tie-ups as proof of exactly that strategy, noting that two of the largest providers are now working with the company.
What the deals actually deliver
Strip away the branding and the partnerships offer two concrete benefits.
The first is cost: The Google programme provides up to $350,000 of cloud credits over two years, alongside training and access to its Vertex AI platform and Gemini models, which should lower Potentially's infrastructure bills during launch and stretch its cash further.
The second is timing: The ByteDance arrangement gives discounted access to the Seed model family, spanning video, image, reasoning and audio, plus early sight of new releases before they reach the wider market.
That early access is the more strategically interesting of the two, since it could let Potentially fold fresh capabilities into its platform before competitors can offer them.
The numbers nobody mentioned
Here is the gap that matters. Neither announcement discloses revenue, customer numbers or the company's cash position.
Cloud credits reduce costs, but they are not sales, and a discount on someone else's models is only valuable if paying customers are using the platform to reach them.
Early access to unreleased models confers an edge only if there is a growing base of users to put that edge in front of.
Without any read on traction, it is impossible to judge whether $350,000 of credits and a head start on new models materially move the company towards profitability, or simply lower the cost of a business still searching for demand.
A crowded lane
The wider context tempers the excitement further.
Aggregating and routing between models has become one of the busier corners of the AI industry, with far larger and better-funded players building similar "use any model" tooling.
Being neutral is a sound strategy, but it is not a rare one, and supplier partnerships are easier to win than customer loyalty.
Potentially has assembled the right pieces on the supply side.
The question its investors still cannot answer is whether anyone is buying what it plans to sell.
