Kabir Jain and Ziming Qiu aren't shy about their claims. Their two-person startup, Inviscid AI, fresh from Y Combinator's Winter 2026 batch, says it can simulate airflow physics in buildings roughly 1000 times faster than the computational fluid dynamics tools engineers have relied on for decades.
The pitch hinges on speed—not for its own sake, but because it unlocks something facility managers rarely attempt: continuous HVAC optimization that adjusts in real time as weather shifts, occupancy fluctuates, and equipment ages. In early March, the San Francisco and Singapore-based company opened its doors with a platform targeting a stubborn problem. Buildings devour about 40% of U.S. energy and roughly 74% of the nation's electricity. HVAC systems alone account for 30-50% of that load, depending on climate and how the building was designed. Small efficiency gains at that scale? They add up fast.
Whether Inviscid can deliver on those numbers in live environments—where theory meets reality—is another question entirely.
The Core Proposition
At its heart, Inviscid AI uses real-time physics simulation to create digital twins of built environments, pulling in streaming data from IoT sensors scattered throughout a building or data center. The result: digital twins that generate continuous spatial fields—airflow patterns, thermal distributions, energy consumption—updated in seconds rather than the hours traditional simulation workflows demand.
The platform taps neural operators and GPU-accelerated numerical solvers, pulling in thousands of sensor feeds simultaneously. It monitors conditions around the clock, autonomously tweaking setpoints based on occupancy patterns, weather forecasts, and equipment status. The company claims energy reductions in the 15-30% range, and says its models maintain 95% accuracy or better.
That's a significant departure from conventional CFD, where modeling a simple vent adjustment might require an overnight batch job. Here, the system recalculates on the fly and adjusts controls to match. The company integrates directly with building management systems, bypassing the cumbersome back-and-forth that typically stalls optimization efforts.
Early Deals and Ecosystem Signals
In February, Inviscid signed a memorandum of understanding with the Madhya Pradesh State Electronics Development Corporation in India—a proof-of-concept deployment at a state data center focused on thermal simulation, hotspot detection, and predictive monitoring. The announcement came during India's AI Impact Summit in New Delhi, a visible moment for a company still working out of a shared office.
Around the same time, Singapore Management University selected Inviscid for its Urban SustaInnovator Cohort 1 program. For a startup straddling two continents, the dual presence in India and Singapore signals intent to build traction in markets where energy costs and cooling demands are acute.
Inviscid's target customers span facility operations teams, data center operators wrestling with cooling and capacity planning, cold storage owners, and building developers. The company is also courting partnerships with digital twin and IoT platform providers—logical allies if the platform is to scale beyond pilot projects.
A Crowded Field, Different Angles
Inviscid is hardly alone in chasing building efficiency. BrainBox AI, which partnered with Trane, reported a 15.8% HVAC energy reduction at 45 Broadway in New York after nearly a year of operation. Honeywell's Forge Sustainability+ platform uses AI-driven optimization across entire building portfolios. In early March, Telefónica Germany announced it was deploying EkkoSense's AI digital twin technology for cooling optimization.
In data centers, Phaidra applies deep reinforcement learning to supervisory control of cooling systems. PassiveLogic, which builds autonomous controls with physics-based digital twins, raised a $74 million Series C last September. The space is noisy, and each player emphasizes a slightly different technical approach.
Perhaps the most frequently cited benchmark remains Google's collaboration with DeepMind back in 2016, which achieved up to 40% reductions in data center cooling energy. Nearly a decade later, the core challenge hasn't budged much: optimizing complex mechanical systems reacting to variables that shift constantly.
What Comes After HVAC
Inviscid's launch materials sketch out ambitions well beyond air conditioning. The founders frame buildings and data centers as a beachhead toward modeling multi-physics systems more broadly—fluids, thermodynamics, electromagnetics, structural dynamics—for virtually any physical infrastructure. That's a long road, and one littered with companies that tried to generalize too quickly.
For now, Inviscid is booking demos with facility operators and data center teams. As a YC Winter 2026 company, the startup has not disclosed details of any priced funding round. Its case studies cite metrics like 240x faster HVAC vent optimization and 600x faster storm surge forecasting—bold figures that lack independent verification or named reference sites. That's not unusual for an early-stage company, but it does mean proof points remain thin.
The question isn't whether the technology works in controlled settings. It's whether Inviscid's speed claims translate into measurable, sustained energy savings once deployed in messy, real-world facilities where sensors fail, occupancy patterns defy prediction, and equipment behaves unpredictably.
Still, the timing might be right. Regulations are tightening, energy costs are volatile, and even mid-single-digit efficiency gains represent meaningful reductions in both carbon emissions and operating expenses. If Jain and Qiu can turn simulation speed into verifiable savings, they'll have a compelling story. If not, they'll join a long list of startups that discovered the gap between lab results and production deployments is wider than it looks on a pitch deck.
