Meta Muse Amazon
AI's Uncomfortable Economics: Grok Hits a Wall, 'Vibe Manufacturing' Takes Off, and Meta Threatens Amazon
Elon Musk acknowledged publicly this weekend that the Grok AI agent is too costly to run at scale, with users hitting usage caps and the current economics of a fully capable consumer AI agent not working at xAI's current pricing. The admission is notable because the AI industry has broadly operated on the assumption that compute costs would decline fast enough to make consumer agents economically viable before capital ran out. If even a well-funded operation like xAI is hitting cost walls, it raises real questions about which AI agent products can survive to profitability.
A parallel story from the hobbyist manufacturing world captures both AI's genuine power and its current limits. Builders are reportedly using GPT-6 Astra to do CAD design work through what they are calling 'vibe manufacturing' — describing desired parts to a chatbot, iterating conversationally on the design, and getting workable output without knowing CAD software. The workflow lowers a historically significant barrier to physical making. But the same users are describing design errors that require multiple physical iterations to correct. The AI handles conceptual approximation; the physics does not accommodate the AI's confidence level. The error cost that is manageable for a hobbyist — a failed 3D print — becomes catastrophic in medical devices or structural applications, which defines the path any serious industrial application must first clear.
Meta's Muse advertising agent, according to a Wall Street Journal report, is posing a genuine competitive threat to Amazon's advertising revenue. The mechanism is that Muse reportedly helps advertisers generate and optimize ad creative directly within Meta's platforms, reducing the need to route spending through Amazon's ad infrastructure. Amazon's advertising segment generated roughly sixty-nine billion dollars in 2025, making it one of the company's highest-margin businesses; even a few diverted percentage points represents a material threat. Muse's competitive edge is not only price — it is friction reduction, combining creative production and targeting optimization in a single interface.
Whether that constitutes an antitrust concern depends on conduct rather than market share. Under Section Two of the Sherman Antitrust Act, having a dominant market position is not itself illegal — regulators must show that a company acquired or maintained monopoly power through exclusionary conduct rather than through genuinely superior products or legitimate business growth. If Meta wins advertising share because Muse is genuinely more useful, that is competition working. If Meta uses its platform access to structurally disadvantage competing tools — degrading performance data for non-Meta creative, for instance — the FTC would have grounds for interest. The central challenge is defining the relevant market: 'digital advertising broadly,' 'social media advertising specifically,' or 'AI-generated ad creative' produce very different assessments of dominance.