The Palo Alto startup emerged from stealth last month with software that predicts GPU performance before workloads run, betting that enterprises will pay to optimize what they already have rather than wait years for new capacity.
Turba Labs announced a $40 million Series A round led by Creandum on October 6, 2026, with participation from Cusp Capital. The financing brings total capital raised to $52 million, including earlier seed funding, since the company's founding in 2023, when co-founders Dr. Patrick Jahnke and Dr. Hans-Juergen Schmidtke began building what they describe as a "digital twin" for AI infrastructure.
The pitch addresses a constraint that has become painfully familiar to enterprise AI teams: GPUs are expensive, power-hungry, and chronically scarce. AI infrastructure investment now runs into the hundreds of billions annually when tallying chips, data centers, and grid upgrades. Data center vacancy in North America has fallen to approximately 1% in recent quarters, making new capacity a multiyear proposition for most companies. Turba Labs wants to double the useful output from machines that enterprises already own.
"We're on a mission to double the world's compute without a single new data center," the company wrote on its website.
How It Works
Turba Labs treats AI workloads and GPU infrastructure as a unified system. The platform builds a calibrated digital twin of a customer's hardware stack, then runs simulations alongside live analytics to predict latency, utilization, and cost before any job executes. Once a workload starts, the software orchestrates routing, batching, replica placement, and power sharing in real time.
The result, Turba Labs argues, is more output per GPU and per watt. In one modeled scenario the company shared, blended cost per token dropped by a factor of 9.4, though the startup was careful to note that figure applied to a specific configuration and should not be taken as a universal benchmark.
"Turba Labs predicts what AI infrastructure will deliver and makes sure it does, workload by workload, in real time," the company said in its announcement. The platform sits between AI workloads and GPUs, making decisions about GPU count, placement, memory allocation, service-level commitments, and power budgets.
Whether that value proposition holds across diverse enterprise environments remains an open question. The company declined to disclose revenue, customer count, or the specifics of its early deployments.
The Founders
Jahnke and Schmidtke both logged decades in enterprise software and hyperscale infrastructure before starting Turba Labs. Jahnke spent 20 years at SAP, where he led predictive maintenance and utilization optimization projects. Schmidtke ran AI infrastructure system engineering at Meta following a career in data center and telecom hardware.
"We're already working with some amazing teams, including SAP," Jahnke wrote on LinkedIn in October. He mentioned Andre Bechtold, SAP's president of industries and chief revenue officer of Industry AI, as an early collaborator. The company confirmed the partnership but would not elaborate on the nature or scale of that engagement.
Turba Labs currently lists between 51 and 200 employees on LinkedIn and is hiring in the Bay Area and Heidelberg, Germany, where it maintains a legal entity called turbalance GmbH. Jahnke said the new capital will fund a commercial organization, expand the customer base, and channel field learnings back into the platform's engineering roadmap.
The startup declined to share its valuation.
A Crowded Field
Turba Labs enters a market already thick with GPU optimization startups. NVIDIA acquired Run:ai, a popular orchestration platform, in late 2024. Intel's Granulate focuses on autonomous runtime optimization with an eye toward energy efficiency. Baseten and OctoAI offer inference orchestration and multi-cloud capacity management.
The distinction Turba Labs draws is one of scope. Most competitors, the company argues, optimize models in isolation. "Other tools only optimize the model. Our software looks at the hardware beneath it," the company wrote on its homepage. Whether that architectural choice translates to measurably better economics for customers will likely determine Turba Labs' long-term success.
The broader backdrop is one of constrained supply. Power availability and grid capacity now shape AI data center development as much as chip availability does. Industry analysts have noted that community opposition and regulatory delays are slowing new builds in key geographies. Cloud providers collectively plan to spend hundreds of billions on AI infrastructure in the coming years, much of it competing for limited power allocations.
What Comes Next
Simon Schmincke, a general partner at Creandum, said on LinkedIn that Turba Labs' platform delivers "more output from every GPU, and more compute from every megawatt." It's the kind of efficiency pitch that resonates in a moment when new capacity takes years to bring online and energy budgets are fixed.
The company is now hiring product managers and expanding its engineering organization on two continents as it looks to scale beyond early enterprise pilots. Whether Turba Labs can turn simulation accuracy into a defensible moat in an increasingly competitive market may depend on how quickly it moves from proof-of-concept to production scale.
For now, the startup has capital and a clear value proposition. The harder work begins.
