Competitive Positioning

The competitive landscape

Vast competes on multiple axes against different sets of competitors:

• Marketplace competitors: RunPod (community side), TensorDock, smaller marketplaces.
• Enterprise neoclouds: CoreWeave, Crusoe, Lambda, Nebius. Different segment but overlap at the boundary.
• Hyperscalers: AWS, GCP, Azure. Compete on the demand side; not really on supply.
• Managed inference: Together.AI, Fireworks, Anyscale, Lepton. Capture the inference workload subset.
• Direct GPU ownership: Some Vast users buy their own GPUs once they hit steady utilization.

Vast is structurally advantaged against some and disadvantaged against others. The honest assessment by axis.

vs RunPod

RunPod is the most direct competitor on the marketplace axis. The two companies serve overlapping customer bases with similar economic structures.

Where Vast wins

• Supply liquidity. Vast has the deeper supply pool by most measures. Customers find what they want faster.
• Price. Vast's larger supply often produces lower prices at the cheap end of the curve.
• Hardware breadth. More consumer cards, more obscure SKUs.
• Operational discipline. Vast's been at this longer; the platform is more battle-tested.

Where RunPod wins

• Developer experience. RunPod's pod templates, serverless offering, and UI are slicker. Lower onboarding friction for non-power-users.
• Two-product strategy. RunPod's Secure Cloud (dedicated, RunPod-owned) gives them a path into segments Vast can't reach — enterprises that want one throat to choke.
• Inference workloads. RunPod's serverless inference offering captures workloads where Vast's raw-VM model is too low-level.

Both companies will likely coexist. They serve overlapping but not identical use cases.

vs TensorDock

TensorDock is a smaller marketplace with a different supply-curation approach. Vast's larger supply gives it an advantage on selection and pricing; TensorDock's curated supply gives slightly more reliability consistency at the average listing.

The two compete most directly for the same long-tail provider base. The strategic question for TensorDock is whether they can build supply density to challenge Vast. So far the gap hasn't closed materially.

vs hyperscalers

Vast doesn't compete with AWS / Azure / GCP on enterprise workloads. But Vast does take some workloads that would otherwise have gone to a hyperscaler at the spot or on-demand boundary.

The hyperscalers' response options:

• Lower GPU pricing. They've done some of this (Azure has been most aggressive). It narrows but doesn't close Vast's price gap.
• Compete on bundled services. Their value proposition becomes "raw GPU price is similar to Vast, but we throw in S3 / IAM / etc." That works for enterprise but doesn't pull Vast's indie / research base.
• Acquire Vast. Possible. The marketplace asset would slot into a hyperscaler portfolio. Cultural friction would be real.
• Ignore Vast. Most likely path. Vast's GMV is small compared to hyperscaler revenue; the segment isn't strategic for them.

The net is: hyperscalers are less of a direct competitive threat to Vast than they are an alternative for customers Vast doesn't well-serve.

vs enterprise neoclouds

CoreWeave, Crusoe, Lambda, Nebius operate in segments Vast doesn't touch. The competition is mostly at the boundary — Series-A/B startups and research-heavy enterprises that could go either way.

The structural truth: enterprise neoclouds and marketplaces are complementary, not directly competing. A given AI company will often use both — Vast for research and experimentation, an enterprise neocloud for production. Each captures different parts of the workflow.

vs managed inference

Together.AI, Fireworks, and similar platforms compete with Vast for inference workloads. The competition is asymmetric:

• Managed inference is per-token; Vast is per-GPU-hour. Different unit economics.
• Managed inference handles model deployment, scaling, optimization. Vast doesn't.
• Managed inference is cheaper for low utilization (you don't pay for idle GPU); Vast is cheaper for high utilization.

The trend has been managed inference winning at most workload sizes. Vast retains the inference workloads where the customer wants direct control or where utilization is high enough that managed pricing is worse.

Vast's moat

What protects Vast's position?

1. The supply corpus

Thousands of providers listed on Vast. A new marketplace entrant would have to recruit comparable supply from scratch — slow and expensive. The two-sided network effect is real.

2. The trust and ranking system

DLPerf, reliability scores, customer ratings. Years of accumulated reputation data. A new entrant starts cold and has to build this from zero.

3. Brand and culture

"Vast" in the indie / research ML community is a known quantity. Word-of-mouth advantage that's hard for newcomers to overcome at the price point Vast operates.

4. Operational discipline and capital efficiency

Vast operates at low headcount with strong margins. Competitors with venture pressure to grow often have worse cost structures and have to push for growth at the expense of unit economics.

5. Provider relationships

Many providers have integrated their fleet operations around Vast specifically — automation, payout flows, etc. Switching costs for established providers are real (not enormous, but real).

Vast's weaknesses

The honest list:

1. Segment ceiling

Already covered in chapter 09. The marketplace can't capture enterprise compliance-bound workloads.

2. Customer churn at scale-up

When indie customers grow into enterprise size, they leave for the segments above. Vast doesn't capture lifetime value on its biggest accounts.

3. No multi-node training fabric

Excluded from the most valuable workload class in modern AI — frontier model training.

4. Limited brand recognition outside the technical community

Vast is a known quantity to people in ML; it's invisible to most enterprise IT decision-makers. The lack of visibility caps category expansion.

5. Supply risks

If a major segment of Vast's supply (post-crypto operators, for example) consolidates or exits, supply density drops. The marketplace doesn't have long-term contracts with providers.

6. Compliance gaps

SOC 2 at the platform level is doable, but extending compliance attestations across heterogeneous supply is structurally hard.

Takeaway

Vast is well-positioned in its segment and structurally bounded by the segment's edges. The moat is real but not absolute. Competitors will keep eating at the edges (RunPod's two-product play; managed inference platforms; enterprise neocloud reserved pricing programs that creep down-market). Vast's strategic challenge is to defend the core while finding adjacent expansions that don't break the marketplace model.

The next chapter covers what's actually happened at Vast recently — the moves, partnerships, and signals from 2025-2026.