Planetary Health March Madness: Efficient Compute takes the crown

For the third year running, RA Capital’s Planetary Health team gathered over fifty investors from across the private and public markets for our annual March Madness event (catch up on the prior years’ madness here). This year’s debates revealed an ecosystem undergoing a transformation, where yesterday’s afterthoughts have become today’s imperatives.

By Kathryn Meng, PhD

PLANETARY HEALTH | SCIENCE | FINANCE

Photo by Kylie Osullivan on Unsplash

March 31, 2026

Planetary Health March Madness returns! Our premise remained the same: a bracket-style head-to-head tournament to predict which sector is likely to have the strongest year in 2026, judged on aggregate capital invested, deal count, exits, and total exit value.

First up: seeding. After substantial debate around seeding strategy last year, we revised our approach. Drawing on operations research literature on tournament design, we structured the bracket to ensure competitive balance across quadrants while delaying confrontations between the strongest contenders until later rounds. This approach, which pairs participants based on equal rank differences rather than traditional 1‑vs-64 seeding, created unpredictable early matchups and ensured no single region of the bracket was predetermined.

Efficient compute dominates its region – but the real story is in critical minerals

Efficient compute won the top left region, a continuation of its momentum from 2025. No one questions the growth story for AI and data centers, or the need for more compute with less energy consumption. But more compute also means more heat, a notable, discussion-generating challenge. Nvidia’s DGX B200, for example, is rated to reject nearly 49,000 BTU per hour of heat – roughly the cooling load of a four-ton commercial AC unit, the kind typically used for a small office or retail space. Nvidia’s next-generation DGX B300 platform is rated slightly higher still. At that scale, cooling is no longer a supporting utility; it becomes core infrastructure. This has created a massive market for cooling technologies, from advanced direct-to-chip cooling (imagine a miniature radiator mounted directly on the processor) to immersion cooling (submerging entire server racks in special non-conductive fluids).

The more substantive debates emerged earlier in the region. Biomining defeated sustainable aviation fuels (SAF) in one of the most contentious matchups. The case for SAF centered on oil prices and policy tailwinds. Biomining proponents leaned heavily on persistent supply deficits and the geopolitical imperatives around domestic and allied-nation mineral sourcing. In February 2026, the U.S. announced over $30B in financing for critical mineral supply chains. Biomining offers a pathway to extract critical minerals from low-grade ore and mining waste at a time when conventional mining alone cannot meet demand.

Metals recycling rode those same trends en route to victories over PFAS remediation, climate fintech, and heavy-duty transport before meeting its match against efficient compute. The IEA estimates that scaling recycling could reduce primary mining demand by ~40% for copper and cobalt and ~25% for lithium and nickel by 2050 (under stated policy scenarios). Recycled critical minerals also carry ~80% lower lifecycle GHGs relative to primary extraction.

Three nations control 86% of critical mineral processing, with China dominating lithium, cobalt, graphite, and rare earth refining. The ability to recover and recycle metals domestically represents both an economic and strategic asset. 

Industrial automation surges, and the SpaceX question looms

Semiconductors/​superconductors won the bottom left region, but that victory came with an asterisk that generated the evening’s most philosophical debate.

Investors highlighted robotics adoption among the region’s formidable also-rans. Warehouse automation is achieving lights-out operations (no workers, no lights needed; fully automated facilities operating in the dark); in agriculture, autonomous tractors and fruit-picking drones are moving from pilot to production; and growing domestic labor shortages are compounding interest across sectors. And design and implementation stages that once took years are now measured in months through robot-as-a-service (RaaS) models, enabling smaller manufacturers and third-party logistics players (3PLs) to deploy automation without prohibitive upfront capital.

After defeating energy retailing, industrial automation and robotics faced nuclear fusion. While our participants projected that fusion will be transformative and have a breakout year in 2026, the consensus was that industrial automation is deployed and delivering measurable returns today, while energy generation from fusion at scale remains on the horizon.

Regional finalist semiconductors and superconductors beat AI for materials discovery before prevailing over industrial automation. Data center build-out and superconductors’ upside in power transmission were too much for the robots: high-capacity power delivery in significantly more compact footprints carried the day.

Meanwhile, space tech defeated GHG measurement, reporting, and verification (MRV) and long-duration energy storage before a showdown with semiconductors and superconductors in the region final. Semi/​supers won, but this is where we encountered key questions: how should we account for the SpaceX IPO, and is space tech planetary tech?

If we included SpaceX, no sector in the bracket could compete on capital metrics. We debated both scenarios: space tech wins if SpaceX is included; semiconductors and superconductors win if it isn’t.

We also asked where there were genuine planetary health plays within space tech. Earth observation satellites enable precision agriculture and climate monitoring. Remote sensing tracks deforestation, methane leaks, and ocean health. Orbital manufacturing could produce materials impossible to create under Earth’s gravity. The opportunity set is real, but it requires disaggregation from SpaceX’s commercial launch and Starlink’s dominance.

But regardless of whether space tech or super/​semi won the region, efficient compute prevailed on monetization clarity. The market needs advanced compute and thermal management today; space tech’s applications, while promising, require more development, especially beyond defense applications.

Substation Technologies’ three-year ascent

If one result captures how the planetary health investment landscape has shifted over three years, it’s this: substation technologies (transformers, power electronics, grid interconnection infrastructure) won the top right region. What was once a quiet corner of industrial infrastructure has become one of the most capital-relevant bottlenecks in planetary health.

When we held our first event in 2024, this category barely registered. This year, it defeated battery materials, which had itself advanced past photonics. The progression tells a story we’ve been tracking closely: the infrastructure required to connect new generation capacity to the grid has become one of the energy transition’s most acute bottlenecks.

The numbers are stark. Combined grid infrastructure requirements for North America and Europe are projected to exceed $1.5 trillion through 2030, driven by the dual pressures of AI-driven load growth and the aging of existing transmission assets. The U.S. electric power sector alone faces unprecedented capital needs, requiring approximately $1.4 trillion in investment through 2030 to fund generation, transmission, and distribution upgrades. At the same time, supply chain constraints are binding: critical grid hardware, particularly large power transformers, is currently experiencing manufacturing lead times averaging more than 2.5 years, creating a significant hardware bottleneck for new capacity. U.S. interconnection queues host over 2.3 terawatts (TW) of active generation and storage capacity — roughly 2× the total currently installed U.S. generation capacity. While renewables accounted for ~93% of all new power capacity additions globally in 2025, deployment is increasingly gated by grid equipment availability, particularly step-up transformers, autotransformers, and related grid infrastructure.

Emerging innovation includes solid-state transformers (SSTs), which use power electronics instead of iron cores, enable dynamic control of voltage and power flow, improving grid flexibility and utilization. They’re gaining traction in data centers and EV charging hubs, where they consolidate multiple pieces of equipment. But they’re not yet widely adopted in substations, where they promise to dramatically improve power flow control, system stability, and asset utilization. The DOE’s Solid State Power Substation Technology Roadmap outlines a vision for modular, standardized converter designs that could reduce component criticality while providing built-in cybersecurity and black start capabilities. But execution timelines remain uncertain, and traditional iron-core transformers will dominate this decade.

Quantum computing wins, but EV charging is the dark horse

Quantum computing won the bottom right region, defeating geothermal in the semifinal and EV charging station equipment and networks in the final.

Geothermal had strong support, fueled by expectations for significant public markets activity. Multiple high-profile geothermal companies are rumored to be preparing IPOs in 2026, including Fervo. Investors in the room were bullish on the potential for enhanced geothermal systems (EGS) to deliver firm, dispatchable power with capacity factors consistently exceeding 90%, offering a carbon-free alternative to fossil-fuel baseload. 

But the region’s genuine surprise was EV charging station equipment and networks. In prior years, this underdog sector struggled due to persistent concerns around monetization and margin compression. But this year it advanced, with genuine excitement in the room.

What changed? Waymo, for one thing. The company’s successful autonomous ride-hailing services (20M+ lifetime fully autonomous trips as of early 2026) may help fundamentally alter the calculus for charging infrastructure. And this is just the tip of the iceberg. The viability of autonomous vehicle networks transforms EV charging from a low-margin consumer amenity into fleet infrastructure with predictable utilization, long-term contracts, and economies of scale; a business model capital can get behind. When vehicles operate 20+ hours per day rather than sitting parked 95% of the time, charging infrastructure economics shift dramatically. Fleet electrification is an investable thesis in a way consumer charging never quite became. Waymo is expected to roll out in Boston imminently, and the consensus was clear: if Waymo can handle Boston roads, the sky’s the limit (we’ll leave that to Wing).

The championship matchup: incrementalism versus the quantum leap

Efficient compute’s victory over quantum computing comes down to timing. Nobody disputed quantum’s long-term potential. Error correction milestones and commercial deployments are advancing. But for a competition grounded in 2026 metrics, the room concluded that efficient compute, underpinning AI infrastructure, data centers, and the digital economy’s energy demands, represents the clearer near-term opportunity for capital deployment and return realization.

Can AI Replace a Room Full of VCs?

In the spirit of experimentation, we ran the bracket through ChatGPT and Claude. We did this cold, with no conversation history or prior context. The results were… uncomfortably prescient. Both AIs independently selected efficient compute as the 2026 champion. 

Before anyone panics about AI automating venture capital: the AIs diverged significantly during the middle rounds. The convergence on efficient compute as our champion likely reflects something quantifiable: the enormity of hyperscaler capex commitments and AI infrastructure buildout. It’s not a subtle call. But the nuance, the why behind the matchups, the shifts in investor sentiment, the geopolitical tailwinds, the three-year narrative arcs, that’s where human pattern recognition still dominates (we hope).

So AIs aren’t quite ready to replace a room full of experienced investors debating over pizza and beer. But we’ll know the real answer when we reconvene in 2027 to score everyone’s brackets, human and machine, against 2026’s actual outcomes. If the AIs nail it, we’ll have some awkward conversations. If they don’t, we’ll toast to the irreplaceable value of lived experience and informed intuition.

What the bracket reveals

Three years in, the bracket continues to be one of our favorite mechanisms for taking the ecosystem’s temperature. The results don’t predict the future, but they reflect where thoughtful, diverse capital is directing attention.

Substation technology’s ascension from afterthought to bracket winner over three cycles tells us more than any individual prediction. The grid, long treated as invisible infrastructure, has become a first-order constraint on the energy transition. Similarly, metals recycling and biomining’s growing importance reflect a fundamental shift from viewing critical minerals as commodities to recognizing them as strategic assets subject to geopolitical competition. And the EV charging story reveals how enabling infrastructure becomes investable when paired with the right business model – in this case, autonomous fleets.

We’ll reconvene at year-end for the retrospective. Whose predictions were championship-caliber? Human or machine? We’ll find out.

What about 2025?

For those curious about how our hive mind performed versus actuals in 2025, here is last year’s 2025 actual performance versus group predictions bracket:

And for a refresh on last year, here’s our summary.