Ten sectors, one pattern. Physical infrastructure, the kind you can touch, inspect, and wait years to permit, has become the binding constraint on the next wave of technology adoption. Software has outrun the pipes, the wires, the reactors, and the packaging lines that it depends on. What we are watching now is the moment when regulatory clearances, capacity milestones, and capital commitments begin to catch up. That catch up is happening across atomic power, grid hardware, cooling systems, chip packaging, drone corridors, and settlement rails, more or less simultaneously, and that simultaneity matters.
The common thread is not a single technology or a single policy. It is the removal of a specific, identifiable bottleneck. A construction permit granted. A criticality milestone reached. A FERC order issued. An FAA certification handed over. Each of these is a gate opening, not a promise of one opening. The sectors below have all passed, or are about to pass, a concrete threshold that was not there six months ago. That is what makes H2 2026 worth watching with unusual attention.
We rank these by the nearness and specificity of the catalyst, combined with our assessment of how much the market has already priced the change in. The top of the list is where the gap between physical reality and market recognition looks widest to us. The bottom is real, but the story is more widely told.
01. Nuclear and Small Modular Reactors
The constraint. NRC licensing timelines measured in years, combined with no commercially operating SMR to point to, kept capital on the sidelines.
What changed. Aalo Atomics achieved initial criticality at Idaho National Laboratory by July 4, 2026, the fourth DOE authorized microreactor to reach that milestone. TerraPower’s Natrium plant then secured its NRC construction permit. Two separate, concrete firsts arrived within weeks of each other.
The window. NRC licensing decisions on multiple SMR designs are expected across H2 2026, any one of which could shift the cost of capital for the entire sector.
Positioned for it.
Oklo ($OKLO): Aurora design is furthest along among publicly traded microreactor developers, with a site use permit at Idaho already in hand.
Kairos Power (private): Fluoride salt cooled design is in NRC construction permit review; a positive outcome would validate the licensing pathway for the field.
GE Vernova ($GEV): BWRX-300 has advanced licensing agreements in multiple countries and benefits from GE’s existing NRC relationships and manufacturing infrastructure.
02. Geothermal and Next Generation Baseload
The constraint. Enhanced geothermal systems lacked proof of commercial scale, and project finance could not get comfortable without a public market comparable.
What changed. Fervo Energy completed a $1.89 billion upsized IPO in May 2026, giving the sector a public price signal and balance sheet. Separately, Zanskar revived a failing New Mexico geothermal plant to full capacity using advanced subsurface modeling and modern drilling techniques, a result reported July 29, 2026, demonstrating that the technology works on brownfield assets, not only greenfield ones.
The window. Fervo Cape Station Phase 1 commercial operations and further EGS project final investment decisions are expected in H2 2026, which would convert the IPO story into an operating revenue story.
Positioned for it.
Fervo Energy ($FERV): Only publicly traded pure play EGS developer, with a power purchase agreement in place and Cape Station construction underway.
Ormat Technologies ($ORA): Profitable, dividend paying operator with an existing geothermal fleet and the engineering capability to apply EGS techniques to its own resource base.
Zanskar (private): Its AI driven subsurface modeling is the methodology that revived the New Mexico plant; the technology could become the underwriting standard for geothermal project finance.
03. Regenerative Medicine and Cell Therapy Manufacturing
The constraint. Manufacturing complexity and opaque FDA chemistry, manufacturing, and controls requirements made commercial scale cell and gene therapy production unreliable and expensive.
What changed. The FDA published CMC flexibility guidance for cell and gene therapies in May 2026, clarifying expectations and lowering the documentation burden for manufacturers. Three regenerative medicine products have already received approval year to date through July 2026, with six more under active review, a pace that reflects the guidance taking hold in real time.
The window. Six pending BLA decisions are expected in H2 2026, against a backdrop of 43 Phase III programs in the pipeline. The approval rate matters less than the pace of decisions, which now has a calendar.
Positioned for it.
Fate Therapeutics ($FATE): iPSC derived cell therapy platform benefits directly from the CMC flexibility guidance, which was written for exactly the kind of manufacturing variability iPSC programs face.
BlueRock Therapeutics (Bayer subsidiary): Cell line manufacturing expertise and Bayer’s regulatory infrastructure position it well for the BLA wave, even as a private entity.
Sarepta Therapeutics ($SRPT): Already approved in gene therapy with ELEVIDYS; additional programs in the pipeline stand to benefit from precedent the company itself helped establish with FDA.
04. Critical Minerals Processing and Refining
The constraint. China’s dominance of rare earth refining meant US buyers had no credible alternative at commercial scale, so domestic investment lacked urgency.
What changed. China’s rare earth exports fell 34% year over year in June 2026, a supply shock large enough to force procurement decisions that were previously deferred. A US executive order on defense critical minerals supply chain consolidation followed, endorsed by Nova Minerals on July 27. The DOE awarded $67 million to Colorado School of Mines and ElementUSA for a Louisiana rare earth processing facility in early July, the first federally funded refinery of its kind.
The window. DOE SPARK program funding selections are due in August 2026; domestic refinery capacity decisions will follow in H2 2026 as the supply shock works through order books.
Positioned for it.
MP Materials ($MP): Only fully integrated rare earth miner and processor operating at scale in the United States today.
Lynas Rare Earths ($LYC.AX): The largest producer outside China, with a US processing facility in development under a DOD contract.
Energy Fuels ($UUUU): Uranium miner that has pivoted to rare earth processing at its White Mesa Mill, giving it a permitted, operating refinery at a moment when permitted capacity is the scarcest input.
05. Semiconductor Packaging and Substrates
The constraint. Advanced packaging capacity, specifically CoWoS and ABF substrates, has been the physical ceiling on AI chip shipment volumes for two years.
What changed. AT&S’s CEO confirmed on July 28, 2026 that CoWoS advanced packaging slots are running 52 to 78 weeks out. Ajinomoto Fine Techno controls 95% of ABF substrate supply with no near term capacity addition sufficient to close the gap. TSMC is targeting 127,000 CoWoS wafers per month by end of 2026, up sharply from earlier capacity, but demand is expanding faster than supply.
The window. NVIDIA’s Rubin architecture packaging ramp begins in Q4 2026, adding a new demand layer on top of Blackwell production that has not yet peaked. Slot scarcity tightens further before it eases.
Positioned for it.
TSMC ($TSM): Controls the CoWoS process and captures the margin premium that comes with being the only qualified supplier at volume.
ASE Technology ($ASX): Largest outsourced semiconductor assembly and test company globally; capacity utilization rises directly with packaging demand that cannot be absorbed by TSMC alone.
AT&S (Vienna: $ATS): High end substrate manufacturer whose own lead times confirm the bottleneck; a beneficiary of any customer willing to pay for allocation certainty.
06. Grid Software and Virtual Power Plants
The constraint. Utilities and grid operators lacked regulatory obligation to compensate distributed resources at market rates, which kept virtual power plant aggregation economics marginal.
What changed. FERC issued six Section 206 show cause orders to all US RTOs and ISOs on June 18, 2026, compelling large load tariff reform or formal justification within 60 days. FERC followed with a grid enhancing technologies technical conference on July 7 and 8, 2026, signaling that the agency intends to move on multiple fronts at once, not sequentially.
The window. RTO and ISO compliance filings are due mid August 2026. FERC Order 2222 full VPP integration milestones arrive Q4 2026, converting regulatory obligation into dispatch reality.
Positioned for it.
AutoGrid (Schneider Electric): Largest installed base of VPP software, positioned to aggregate compliance driven distributed resource enrollments as utilities respond to FERC pressure.
Stem ($STEM): Athena AI platform manages battery dispatch across commercial and industrial sites; Order 2222 compliance makes those assets worth more inside a VPP structure.
Tesla ($TSLA) Autobidder: Real time energy trading platform with direct integration into Powerwall and Megapack deployments; the only operator with both the software and the hardware fleet at scale.
07. Transmission and Grid Hardware
The constraint. Transmission congestion and aging infrastructure were known problems, but capital allocation required both federal direction and utility commitment to move at speed.
What changed. A DOE draft report published July 10, 2026 showed transmission congestion cost $12 billion in wholesale power in 2024 alone, a number concrete enough to anchor regulatory urgency. EEI confirmed $208 billion committed for US grid infrastructure in 2025 alone. DOE SPARK reconductoring awards are expected in August 2026, the first tranche of a program designed to increase capacity on existing corridors without new right of way.
The window. SPARK award selections in August 2026 will name specific projects and contractors. PJM governance reforms are due from the July 23 FERC conference, which would accelerate interconnection queue processing in the largest US grid region.
Positioned for it.
GE Vernova ($GEV): Grid equipment backlog is already multi year; transformer and switchgear manufacturing benefits directly from the $208 billion commitment.
Hitachi Energy (Hitachi: $6501.T): High voltage direct current technology is essential for long distance transmission corridors; order book visibility extends well into 2028.
Quanta Services ($PWR): The largest transmission construction contractor in the US; every dollar of committed grid capital eventually flows through an engineering and construction firm.
08. Drone Logistics and Advanced Air Mobility
The constraint. No major consumer platform had its own FAA certified drone delivery operation, which left the market fragmented among specialists with limited distribution reach.
What changed. DoorDash received FAA Part 135 air carrier certification on July 29, 2026 and launched DoorDash Air, its own in house drone delivery network. It became the eighth US operator to earn the certification. The significance is the platform: DoorDash brings 37 million customers and existing merchant relationships to a certified air operation, a combination no prior certificant had.
The window. DoorDash Air pilot operations and network rollout details are expected in H2 2026. The FAA’s eIPP program is targeting commercial adoption across 26 states by late 2026, which would open corridor access for all certified operators simultaneously.
Positioned for it.
DoorDash ($DASH): Certified operator with the largest consumer delivery network in the US; the distribution advantage compounds as route density improves unit economics.
Joby Aviation ($JOBY): FAA type certification process is the most advanced among eVTOL developers; an air taxi certification would follow the same regulatory pathway that DoorDash just demonstrated works.
Archer Aviation ($ACHR): Midnight aircraft in FAA certification process; United Airlines partnership provides commercial launch infrastructure if certification arrives on schedule.
09. Thermal Management and Data Center Cooling
The constraint. Air cooling cannot handle the thermal density of GPU clusters running at 300 watts per chip and above, but liquid cooling supply chains were too fragmented to support hyperscaler procurement at scale.
What changed. Eaton closed its $9.5 billion acquisition of Boyd Thermal and Ecolab paid $4.75 billion for CoolIT, consolidating coolant distribution unit supply into two well capitalized platforms. Modine announced a US manufacturing expansion for Airedale data center cooling products in July 2026, adding domestic capacity at a moment when supply chain localization is a procurement requirement for many hyperscalers.
The window. 23 gigawatts of data centers are under global construction now. Hyperscaler capital expenditure commitments of $750 billion for 2026 are already driving CDU order books into 2027; the constraint is manufacturing throughput, not demand.
Positioned for it.
Vertiv ($VRT): Largest independent thermal management company globally, with a cooling backlog that has grown for eight consecutive quarters.
Modine Manufacturing ($MOD): US manufacturing expansion positions it as the domestic supplier of record for customers who cannot use foreign sourced cooling equipment under current procurement rules.
nVent Electric ($NVT): Precision cooling and liquid cooling enclosures address the rack level thermal problem; revenue mix is shifting toward data center at the fastest pace in the company’s history.
10. Tokenization and Market Infrastructure
The constraint. Settlement infrastructure remained siloed from on chain activity, which meant tokenized assets could not interact with traditional securities workflows at any meaningful scale.
What changed. DTCC began limited production tokenization trades in July 2026, with more than 50 institutions including BlackRock and Goldman Sachs participating. The SEC approved Nasdaq rule changes in March 2026 and NYSE changes in April 2026 to accommodate tokenized securities. On chain real world asset value reached $33.5 billion, a fourfold increase from early 2025. These are not experiments. They are live trades on production infrastructure.
The window. DTCC targets a full tokenization service launch in October 2026. The SEC’s crypto safe harbor package is expected to be finalized in H2 2026, which would establish the legal framework for the asset class at scale.
Positioned for it.
Coinbase ($COIN): Base blockchain and custody infrastructure are already used by institutions participating in DTCC pilots; a full launch expands that relationship from optional to essential.
Ondo Finance (private): Largest tokenized US Treasury product by assets under management; the on chain RWA growth since early 2025 is partly its own AUM.
Securitize (private): Transfer agent and tokenization platform for BlackRock’s BUIDL fund; sits at the intersection of the incumbent asset management world and the on chain settlement layer that DTCC is building toward.
What would change our mind
On nuclear, a negative NRC licensing decision on any of the SMR designs under review in H2 2026 would reset timelines by at least two years and damage confidence across the sector regardless of the specific applicant. On geothermal, a Fervo Cape Station delay past Q1 2027 would re open the question of whether EGS projects can hold their construction schedules at commercial scale. On cell therapy, an FDA complete response letter on any of the six pending BLAs citing manufacturing deficiencies would signal that the May guidance has not yet changed the agency’s underlying standards in practice.
On critical minerals, a partial reversal of China’s export restrictions, or a diplomatic agreement that softens the constraint, would reduce the urgency behind domestic refinery investment. On semiconductor packaging, a demand correction in AI infrastructure spending, which we think is unlikely in the next two quarters but not impossible in 2027, would lengthen lead times mechanically without any supply addition. On grid software and transmission, a FERC leadership change that slows Section 206 enforcement would extend the compliance timeline and defer the Order 2222 revenue moment.
On drones, a category wide FAA operational restriction following any high profile incident would pause the eIPP program and reset the certification queue. On cooling, a meaningful reduction in hyperscaler capital expenditure guidance, from any of the four largest operators, would compress CDU order books faster than the sector currently expects. On tokenization, an SEC enforcement action against a DTCC participant that challenges the legal standing of on chain settlement would cloud the October launch. We watch all of these. None has happened yet.