Three of the biggest names in agent infrastructure shipped product announcements in the last two weeks. Meta launched its Enterprise Platform on September 28, bundling the Muse agent, Meta Business Agent, Muse API, and Muse Code into a single enterprise play. Apple’s HomePad – a smart home hub built around its rebuilt Siri AI – leaks point to an October 13 announcement. xAI’s GrokBot has been available through subscription bundles since early September.
None of them have published a price for their agent platform.
Not “contact sales for custom pricing” with a public reference rate. Not a tier page with a “call us” button on the enterprise row. Just silence – or, in GrokBot’s case, a bundle structure so fragmented that the standalone cost of the agent layer is impossible to extract.
That’s not a coincidence. It’s a signal.
The stack consolidated. Now the competitive variable is pricing.
Four weeks ago, the agent infrastructure market looked like a protocol war. Now the protocols are settling. MCP won the interoperability standard. A2A is finding its lane for agent-to-agent communication. The orchestration layer – the routing, tool selection, and guardrails that determine what an agent actually does – is being decomposed into specialized small models (see: the decision layer unbundling).
What remains is the pricing question. And three major players are simultaneously declining to answer it.
Meta’s Enterprise Platform has no published pricing, no SLAs, no enterprise contracts, and no named launch customers. The per-token rate for Meta Business Agent via WhatsApp Business Platform API is public ($2.00 per million tokens), but that’s a component price, not a platform price. For the full bundle – Muse agent, Muse API, Muse Code – buyers go through Meta sales. Enterprise procurement teams evaluating the platform against Claude Enterprise ($20/seat/month) or ChatGPT Enterprise (custom via sales) have no public reference point.
Apple’s HomePad is a hardware play first, with a rumored ~$350 price point for the device. But the Siri AI agent layer – the part that matters for anyone building on or integrating with Apple’s agent infrastructure – has no published pricing. And the device is personal-only. There’s no enterprise tier, no API, no developer program for Siri-as-agent-platform. Apple is shipping a consumer entry point and leaving the enterprise half of the market entirely open.
GrokBot is the most revealing case. xAI’s own Business and Enterprise tiers don’t include it. To access GrokBot, enterprise buyers must either subscribe to SuperGrok ($30–$300/month, consumer-oriented tiers with weekly usage allowances) or go through Cursor Enterprise (custom pricing). The standalone cost of the agent layer is buried inside bundles designed for other purposes. That’s not a pricing strategy – it’s a pricing tell. When you can’t extract the cost of the component you’re evaluating, the vendor hasn’t decided what it’s worth yet.
Why the vacuum matters
When one platform withholds pricing, it’s a sales strategy. When three platforms in the same market segment withhold pricing simultaneously, it’s a market structure signal. Either the players are waiting for someone else to commit first – a coordination problem – or the market hasn’t settled on what “agent platform” is actually worth to an enterprise buyer.
The procurement consequence is real. Enterprise IT teams need total cost of ownership to justify adoption. They need to compare apples to apples across Meta, Apple, xAI, and the established players – Anthropic, OpenAI, Google. Without published pricing, those comparisons can’t happen. The result is slower adoption, more pilots, and more “wait and see” postures from exactly the buyers these platforms need to reach scale.
The irony is that the infrastructure layer – the protocols, the orchestration, the runtime – has never been more accessible. Open-source models, open protocols, and commodity compute have driven the cost of the stack down. But the platform layer, where those components are bundled into something an enterprise can buy, deploy, and manage, remains opaque.
The agent platform market is consolidating around three or four major players. The next competitive variable isn’t capability – it’s what the capability costs. And right now, nobody is showing their cards.