Modal closes $355M Series C at $4.65B valuation

Modal Labs announced a $355 million Series C on May 21, 2026, at a $4.65 billion post-money valuation. General Catalyst and Redpoint led, with Menlo, Bain Capital Ventures, and Accel joining as new investors and every existing investor following on. The size of the round, the speed of the markup (the company was valued at roughly $1.1 billion as recently as October 2025), and the underlying revenue trajectory make this the largest financing in the serverless-GPU category to date.

What the round looks like

The deal closed in two tranches. The first priced at a $2.5 billion post-money valuation; as more investors came in, Modal opened a second tranche at the final $4.65 billion mark. Total capital raised to date now sits at roughly $465 million, on top of the $87 million Series B that Lux Capital led in October 2025.

General Catalyst and Redpoint are the named co-leads. Menlo Ventures, Bain Capital Ventures, and Accel are the new entrants. Modal has not disclosed the exact allocation between funds, but the structure (“doubled down” existing investors plus three brand-name additions) is the shape of a round priced by inbound demand rather than a syndicated pitch.

The growth that pulled it in

Modal’s annualized revenue ran to roughly $300 million at the time of the announcement, up from about $60 million in September 2025. That is a fivefold expansion in roughly six months, a pace that explains the 4.2x valuation step from the Series B.

Two operating numbers from Modal’s own announcement frame the shape of the workload mix. The platform has launched over 1 billion Sandboxes (Modal’s per-task isolated execution primitive), and Sandboxes now account for more than one-third of revenue. The customer roster Modal disclosed spans agent-platform builders (Cognition, Decagon, Applied Compute), AI-native products (Suno, Reducto, Chai Discovery), and large incumbents running production workloads (DoorDash, Ramp). Modal also says it now integrates with 13 cloud providers, up from 5 a year earlier.

Where the money goes

Modal’s announcement names three priorities for the capital, each mapped to a concrete platform investment:

1. Low-latency inference at scale. Continued work on serving primitives, observability, and what Modal calls the “open inference stack”. For teams using Modal as a hosting layer for custom inference, this is the bucket that affects steady-state serving cost and tail latency.
2. Collapsing the training and inference loop. Modal is investing in RL infrastructure and broader coverage of the full model training lifecycle. The framing matters: Modal has been a strong fit for inference and fine-tuning but a weaker default for full pre-training and large-scale RL. The Series C is partly the bet that the same Python-native developer surface can win at training too.
3. The compute layer for agents. Sandbox capacity for millions of parallel executions, plus granular RBAC. Sandboxes already drive a third of revenue, and the priority list suggests Modal expects that share to grow as agent platforms (its own customers among them) scale.

Hiring is the operational lever. Modal plans to expand engineering across New York (its headquarters), San Francisco, and Stockholm.

The competitive picture

The serverless-GPU and AI-infrastructure category has clear shape heading into the second half of 2026, and Modal’s round redraws the leaderboard. Three reference points worth comparing.

• 
  Replicate remains the cleanest “model as API” option in the category. A user posts a request to a model endpoint and Replicate handles the rest. Modal’s wedge is the opposite shape: write Python, deploy that Python to a GPU, run inference inside your own process. Modal’s round signals that the “custom code on GPU” pattern is winning the larger budget category, even though both shapes will coexist.
• 
  Together and Fireworks compete most directly with Modal on hosted inference, but their default is a managed model API. Modal’s deeper bet on the training-and-inference loop puts it in a different lane: a runtime for whole pipelines, not a menu of inference endpoints.
• 
  Fly.io and Runpod compete on the “rent a GPU by the second” framing. Modal sits one layer up: opinionated Python primitives (Function, Sandbox, Volume, Cron) instead of bare containers. The pitch to a builder is fewer moving parts; the pitch to a CFO is more dollars per workload.

For Modal customers, the practical near-term read is that the runway question is settled. The platform is unlikely to be acquired or wound down on a short horizon, the price-per-GPU-second card is unlikely to move much (Modal has not signalled any pricing changes), and the platform investments above are the things to watch for over the next two quarters.

Follow-up

The next test is whether the disclosed numbers and stated priorities hold up over a steady-state quarter. Over the next 30 days, the key signals to watch are:

• Shipped product changes against the three priorities (low-latency inference, training-plus-inference loop, sandbox capacity). What actually lands as platform work versus what stays announcement-only.
• Pricing or tier adjustments. Modal did not signal pricing changes at announcement; any movement after a round of this size is significant.
• Revenue trajectory. Whether the $300M ARR figure holds or accelerates in subsequent disclosures.

Stackmaven’s follow-up coverage lands on or around June 23, 2026.