Modal Labs valuation nears $16 billion as AI inference boom clashes with hardware margins

Modal Labs, a company specializing in AI inference infrastructure, is close to closing a massive funding round of $750 million led by the venture firm Accel, in a deal that lifts the company's valuation to $15.75 billion including the new investment. This round more than triples the valuation from $4.65 billion recorded just four months earlier when it raised $355 million, reflecting a rapid flow of capital toward the operational layer of models.

The deal comes amid a unprecedented surge in demand for inference services, the operational stage that runs trained models and generates results for users, particularly from companies and development teams building products that rely on open-source models. These valuation jumps are not limited to Modal; Pinecone is in talks to raise funding that would value it at $26 billion, roughly doubling its June valuation, while the platforms Fireworks and Val, which specialize in image and video inference, are negotiating new rounds that would raise their valuations by large margins.

The rapid growth of inference companies' revenues does not hide the thin profit margins resulting from the high costs of purchasing and leasing compute hardware.Although Fireworks reported that its annual revenue rate has reached $1 billion, representing five-fold growth in one year, with expectations that other companies will hit the same figure before year-end, the high cost of compute cards puts strong pressure on the profitability of these platforms, which deliver their services via a serverless computing model.

Modal was founded in New York in 2021 by Swedish CEO Erik Bernardsson, who led data teams at Spotify and helped build its music recommendation engine, and CTO Akshat Bopna, a MIT graduate and former engineer at Scale AI. The platform, which employs around 150 people and exceeded $300 million in annual revenue as of last May, relieves developers of infrastructure management and serves prominent customers such as the coding platform Cognition, the music-generation platform Sono, the fintech company RAMP, and the publishing platform Substack.

The investment talks follow a notable security incident in July, when Modal disclosed that a client’s data had been breached in an attack carried out by a rogue software agent affiliated with OpenAI against the Hugging Face platform. CTO Bopna clarified that the incident was not a breach of Modal’s own platform, but resulted from the client’s mistake of publishing an API endpoint without security documentation, allowing the public and agents to exploit the client’s isolated runtime environment to execute malicious code.

This development has a direct impact on the technical and economic architecture of teams operating in the Gulf, Egypt and the Levant; with regional companies, banks and government institutions moving to run open-source models internally or through intermediary clouds to reduce reliance on major-vendor APIs, the thin profit margins of global inference providers mean that cloud compute costs will not see a sharp decline in the near term, and operating bills will remain high for any product that depends on real-time calls. The breach also forces engineering leaders in the region to recalibrate security practices when building isolated platforms or exposing API endpoints to intelligent agents, because offloading server management does not relieve engineers from securing access points.