Newsletter · · Ashutosh Agarwal

Tariffs Turn Into a Tailwind While Reshoring Runs Into a Power and Materials Wall - Trade War, Tariffs & Reshoring - Week of July 13, 2026

Trade, tariffs, reshoring and automation newsletter for the week of July 13, 2026. Net tariff revenue is tiny and refunds are flowing as the legal machinery shifts to Section 301, USMCA moves to an annual review, the first pure-play humanoid-robot company goes public, and the binding constraint on America's industrial ambitions turns out to be power, transformers and 60 obscure gases rather than trade policy.

Trade War, Tariffs & Reshoring

Week of July 13, 2026: Tariffs Turn Into a Tailwind While Reshoring Runs Into a Power and Materials Wall

A weekly read on tariffs, reshoring, factory automation, and the physical build-out of the economy, drawn entirely from the past week's podcasts. We separate the people who actually run these businesses (operators and insiders) from the people who talk about them for a living (analysts and pundits), because the two are worth very different amounts.

The one-line version

Two years into the tariff era, the trade story has quietly flipped from "headwind" to "tailwind": tariff revenue is tiny, refunds are flowing, and the legal machinery just changed underneath everyone's feet. Meanwhile the real action has moved from the customs house to the factory floor and the power grid: the first pure-play humanoid-robot company is going public, a carmaker is quietly rehiring the humans it replaced with AI, and the single biggest constraint on America's industrial ambitions turns out to be electricity, transformers, and 60 obscure gases, not tariffs.

1. The trade desk: tariffs are now a stimulus, and USMCA is on the clock

Start with the number that reframes everything. On The Real Eisman Playbook, political strategist Dan Clifton walked through where U.S. trade policy actually stands, and the figures are startling for anyone who still pictures tariffs as a wall of cost. The U.S. Treasury has collected just $35 billion of net tariff revenue year-to-date ("an amazing number. Like, it is nothing," as Clifton put it) while the government has already handed back $23 billion in tariff refunds in a single month. The effective tariff rate has fallen from 11% last October to about 7% now, a 400-basis-point drop, as supply chains adjust and the courts intervene. His punchline for investors: tariffs have "moved from a headwind on tariffs to a tailwind," and this year the government handed out $75 billion of corporate tax cuts (100% expensing of capital goods and R&D) against that $35 billion of tariff income, so the net effect on the economy is stimulus, not drag. (Analyst/pundit voice.) The Real Eisman Playbook

The legal plumbing behind this matters, because it is changing right now. Clifton's summary: the Supreme Court threw out the President's emergency (IEEPA) tariffs in February, and the replacement, built on Section 301 authority, which is on firmer legal ground, "will start to come into effect in July." In plain terms, the administration is rebuilding the same tariff structure through a new, more court-proof legal door, and the refunds are the one-time cost of the old door being slammed shut. Why this matters: a company modeling landed costs off last year's rates is now working from stale numbers in both directions.

The dominant policy story of the week was USMCA, the North American trade pact, and it is genuinely up in the air. Several podcasts converged on the same facts. The U.S. did not agree to renew USMCA in its current form at the July 1 joint review; the deal stays in effect (no new tariffs snapped on automatically), but the review produced no renewal, and the structure has quietly changed from a six-year review cycle to an annual one, meaning the whole agreement now comes up for renegotiation every single year. The next concrete date: the U.S. meets Mexico the week of July 20 for a third round of bilateral talks. The Hard Country · Selling on Giants

The most substantive breakdown came from The Trade Guys, which hosted Diego Marroquín Bitar of CSIS (a trade think tank). His analysis is the kind of thing to internalize before touching any auto or Mexico-exposed name:

  • The auto content fight is the crux. The Trump position is to insert a brand-new 50% U.S.-content requirement into the rules of origin (there is none today) and to raise the North American content threshold from 75% to 82%+. The problem: roughly 40% of a Mexican-assembled vehicle is currently U.S. content, well below the proposed 50% floor. So even a long phase-in leaves the industry scrambling.
  • Here's the counterintuitive part that hurts U.S. producers. Even when a vehicle complies with USMCA, the non-U.S. content still pays a 25% tariff. Run the math on that 60% non-U.S. share and Mexican-built vehicles face an effective 10-15% tariff rate, sometimes higher than what Japanese, Korean, or European finished cars pay (fixed at 15% under their own deals). Parts cross the border multiple times for value-added work, "so the pain compounds." Marroquín's blunt read: raising your own input costs is "self-inflicting damage" that lets China and the EU out-compete a less-efficient North American industry.
  • Uncertainty is the real tax. The administration's theory is that not knowing the rules pushes companies to invest in the U.S. Marroquín disagrees flatly: "I see uncertainty as a tax on investment… you perhaps are going to hire less people. If you're a foreign investor, you might try to set up production elsewhere." His concrete example: Toyota was going to build the Tacoma in Mexico and just moved it to its San Antonio plant, "even a year ago would have been a no-brainer to build it" in Mexico.
  • Canada vs. Mexico have played it very differently. Mexico's President Sheinbaum leaned into a warm rapport with Trump (an asset with Mexican markets); Canada found closeness politically "taboo" amid "51st state" rhetoric and refused to negotiate until Section 232 steel/aluminum tariffs were gone. Canada has now accepted those 232 tariffs are "very likely not going to go away" and is finally starting negotiation rounds, though no dates are set. Marroquín thinks the endgame, if it happens, is a more restrictive USMCA (higher rules of origin, plus new "economic security" measures to screen Chinese investment), essentially a "Fortress North America." (Expert/pundit voice, but the most operator-relevant analysis of the week.) The Trade Guys

The supply-chain view underlines the fragility. On Supply Chain Secrets, the host stressed how intertwined U.S.-Canada auto manufacturing is (parts "move back and forth the border multiple times," with value added in each country), so stepping out of USMCA "would have significant [repercussions] for those supply chains." Their sober take on the administration's mixed signals: "it might mean that nothing changes in USMCA and this is all just talk. Or it might be that whatever trade relations are in place… could change from one day to the next." (Operator-adjacent/practitioner voice.) Supply Chain Secrets

The practical playbook for brands came from Selling on Giants, an e-commerce operators' show, which gave exactly the advice a sourcing manager needs: "monitor, not panic." Brands sourcing from Mexico or Canada should "understand their country of origin exposure, review supplier options, and model how changes could affect landed cost. But no one should move production based on a headline before the rules actually change." The bigger lesson they drew: the great migration out of China into "Mexico, Vietnam, India" reduced concentration risk but added new trade, compliance, and labor risks, "diversification reduces concentration risk, but only when the brand actually understands the new risks it is taking on." (Operator voice.) Selling on Giants

Commodity corners where tariffs are the whole story

  • Copper: the pending Section 232 decision is the catalyst, not supply. On J.P. Morgan's At Any Rate, the commodities team argued that for the rest of 2026, "the key catalyst for [copper] prices is policy, not balances." The Commerce Department's Section 232 copper review was due back to Trump by June 30, and markets are still waiting. Copper is trading around $13,500/ton, a level they say already bakes in "embedded tariff uncertainty," a tug-of-war where the U.S. is pulling metal in while China (structurally short and needing imports) is forced to bid up its buying floor. Their base case: the administration comes out with "some sort of escalatory tariff scenario in the coming weeks," pushing LME copper toward $15,000/ton in H2 2026. At Any Rate
  • Lumber: Canadian duties are about to drop. On The Lumber Word, veteran analyst Russ Taylor flagged that new anti-dumping duties on Canadian softwood are set to fall from 35% to 25%, "probably in August, it could be October," which, layered with the 10% tariff, takes the all-in rate from ~45% down to ~35%. The bigger structural point from the traders: North America has permanently cut capacity, "we've dropped… six billion board feet in the last three years" of closures, so U.S. mills (with no duties) are earning "ungodly amounts of money," $100-150 per thousand board feet, while Canadian mills are only now clawing back to breakeven. (Operator/practitioner voices, mill traders.) The Lumber Word
  • China agriculture: a handshake, not a purchase order. On Grain Markets and Other Stuff, the takeaway was skepticism. China's Ministry of Commerce said the U.S. and China "agreed in principle to reduce tariffs on agricultural products under a reciprocal… framework," with no timeline. A May 17 White House fact sheet targets $17 billion/year in non-soybean U.S. ag purchases. But traders "are looking for actual sales, not just broad diplomatic statements," new-crop Chinese soybean buying is still stuck around a token 200,000 metric tons. The precedent cuts both ways: China did deliver on a prior 12-million-tonne old-crop soybean commitment (actually hitting ~101% of it), so the framework isn't hollow, it just hasn't shown up as corn yet. (Practitioner/market voice.) Grain Markets and Other Stuff
  • A vivid reminder of who eats the tariff. On The Foundr Podcast, the founder of Pur Gum (a Swiss-made chewing-gum brand doing ~$250M in revenue) described absorbing a 39% U.S. import tariff plus a ~20% currency swing, "north of 50% on top of the cost of goods." Rather than raise prices or drop retail listings, the company ran negative margins on product to protect relationships during its biggest growth year (30%+): "It was the first time that we were running negative margin on product… I ate it." Switzerland, he noted with disbelief, got "the most expensive tariff in the world, second only to Myanmar." His survival lever was geographic diversification (double-digit share in Canada, #1 on Amazon Europe, sales in Australia): "your whole margin blends down, but you're able to stay alive." (Operator voice.) The Foundr Podcast

2. Reshoring: the fabs came home, but the stuff that feeds them didn't

The most important reshoring insight of the week is a warning, and it came from ChinaTalk, in a deep conversation with a semiconductor-gases industry veteran. The gist: the CHIPS Act worked at the top layer and failed at the bottom. It successfully got TSMC to build in Arizona and Samsung to invest in the U.S., so there will be more domestic chip fabrication. But it did nothing about the layer underneath: the 60+ specialized chemicals and gases every fab needs. As the guest put it, all these domestic production requirements "are being announced, but without any of the supply chain underneath it," leaving new U.S. fabs facing "a total reliance on the import of almost every chemical required."

The vulnerability isn't about cost (gases are only about 10% of a chip's bill of materials), it's about single points of failure. "Which one of those 60 can you live without and still manufacture your products? … The answer is zero. None of them." And the punchline that should worry anyone long the Western chip supply chain: "Taiwan is 100% reliant on Chinese supply chains today. 100% reliant." If China restricted exports of a single cleaning gas like NF3, "the Taiwanese fabs would shut down."

Here's why China got there and the U.S. didn't. China's "Big Fund" (now roughly $120 billion invested) built the entire ecosystem at once, managed region-by-region, with provinces competing to build the same materials. The result was massive overcapacity: one Chinese province produces 55,000 tons of NF3 a year against domestic demand of only ~8,000 tons, so it now sells the surplus into Taiwan, Korea, and the U.S. at near-zero incremental cost. China also controls 70% of the world's fluorspar (the rock upstream of the whole fluorine chemical family). The one Chinese weak spot the guest flagged: helium, where geology favors the U.S. and LNG producers. The investable read: America's fab reshoring is real but structurally incomplete, and the materials layer is where the leverage, and the risk, actually sits. (Operator/insider voice, industry practitioner.) ChinaTalk

Governments are now buying equity in the supply chain, not just subsidizing it. On The Canadian Investor, the hosts covered Ottawa's first deal under its new Critical Minerals Accelerator: a strategic investment of up to C$400 million (via the Canada Growth Fund) into Teck Resources, backing expansion at Teck's Trail smelting-and-refining complex in B.C., one of the world's largest integrated polymetallic operations, producing minerals critical to defense and semiconductors. The deal even gives Ottawa rights over future production of certain critical minerals. The hosts explicitly linked it to a global pattern: the U.S. has taken stakes in Trilogy Metals, Lithium Americas, Intel, and U.S. Steel. Their thesis: "you just can't assume… that country will continue supplying you next year, 3 years, 5 years down the line," so governments are securing sovereign, in-jurisdiction supply, "the first of quite a few." (Analyst/investor-commentary voice.) The Canadian Investor

3. The factory floor: robots go public, and a carmaker walks back AI

This was the week's richest theme, and it split cleanly into two stories that seem to contradict each other but don't.

Story A: the first real humanoid-robot business is going public (operator)

The marquee event: Agility Robotics, the Oregon maker of the bipedal robot Digit (already working in Amazon warehouses and on Toyota assembly lines), is going public via a SPAC merger with Churchill Capital that values it at ~$2.5 billion and raises $620 million in gross proceeds, the largest capital raise in humanoid-robotics history and the first pure-play humanoid company to list. CEO Peggy Johnson (ex-Magic Leap, ex-Microsoft) laid out the business on StrictlyVC Download, and the specifics are what make it credible rather than hype:

  • $300 million in booked revenue, which she confirmed represents roughly 1,000 robots deployed on multi-year "robots-as-a-service" contracts. A pipeline of 30+ interested companies, all vetted; Agility won't even admit a customer to its program unless they have concrete deployment plans.
  • On valuation froth, she was refreshingly disciplined. Asked about Figure AI's $39 billion valuation, she said it's "really hard to even guess at what the right valuation is," and defended Agility's own $2.5B as "spot on" for "the entry point of humanoids," which today means simple material handling, not the home, not general-purpose labor.
  • Unit economics: roughly 75% of parts are made in the U.S. (a deliberate supply-chain choice, "good to have the parts right here in the U.S."), the company is vertically integrated, on its fifth generation of robot, and, notably, says it is not losing money on robot sales. The pricing model is a monthly RaaS fee pitched below "fully burdened human rate," so the robot delivers "ROI on day one" for a customer who can't find or keep a human. The scale robot (v5, out early 2027) can work outside a safety cage, which she says blows the addressable market wide open.
  • The demand driver she keeps returning to is not cost-cutting, it's a labor shortage that has become "almost existential" for customers: "a retiring workforce… younger people who don't want these jobs that are more manual in nature." (Operator/insider voice, CEO.) StrictlyVC Download

A second operator story rounds out the "robots that actually work" picture. On The Robot Report, the CEO of Path Robotics described using AI to crack robotic welding, a 50-year-old application that "plateaued" because traditional welding robots couldn't handle the part-to-part variability of small manufacturers. Path built a "weld world model" trained with reinforcement learning (the same technique behind superhuman game-playing AI) that does real-time visual correction as it welds. Having started with tier-2/3 shops, they now deploy into Fortune 500 manufacturers. The forward-looking piece: their new product Rove puts a welding torch on a Boston Dynamics quadruped robot so welding can go to the part instead of the part coming to the robot, aimed squarely at shipbuilding (aircraft carriers, submarines) and energy/AI data-center infrastructure, where huge structures are welded outside any factory. First customer deliveries in early 2027; they planned to ship just 50 units but are "already fully at pre-sales" and may raise the number. This is a concrete, reshoring-adjacent tell: the demand is coming from U.S. shipyards and data-center builders who physically cannot find enough welders. (Operator/insider voice, CEO.) The Robot Report

The market backdrop, for scale: on Venture In The South, the hosts cited a PwC analysis (July 1) putting the global physical-AI market (robots, autonomous vehicles, smart infrastructure) at $430 billion by 2030 and ~$1.6 trillion by 2040, with automotive at ~40% today and heavy industrial automation and warehouse logistics close behind. They also flagged the safety-certification bottleneck being addressed head-on: NVIDIA launched "Halos for Robots", billed as the first full-stack safety system for physical AI, drawing on 18,000+ engineering-years of self-driving data, and Agility was the first to adopt it for Digit in Amazon and GXO warehouses. (Analyst/venture-commentary voice.) Venture In The South

Story B: the AI-labor honeymoon is ending (the counter-narrative)

Against all that momentum, the week's most quoted cautionary tale was Ford quietly rehiring the humans it had tried to replace with AI. Discussed on Valuetainment (citing Bloomberg/BBC reporting), the facts: Ford had rolled out 900 AI-powered cameras across plants to catch quality issues, then rehired more than 300 veteran quality inspectors after the automated checks failed to match human skill. The money quote from Ford's VP of Vehicle Hardware Engineering, Charles Poon: "Mistakenly, we thought that by just introducing AI and ingesting the design requirements that we had, that would produce a high-quality product." He noted many veteran technicians had "left the company before their knowledge could be used to improve" the tech. The hosts stacked it alongside Klarna reversing its AI customer-support push and Duolingo softening its "AI-first" stance, a broader pattern of companies discovering AI's limits in the physical, high-stakes real world. (Operator-reported / pundit-framed.) Valuetainment

How to reconcile A and B? They aren't actually in conflict. The Ford story is about AI replacing judgment (quality inspection) and failing. The Agility and Path stories are about robots doing physical work no one will hire for (moving 50-pound totes, welding submarine hulls) and succeeding on ROI. The investable line runs between "AI as a substitute for skilled human judgment" (still oversold) and "robots filling jobs that are genuinely unfilled" (real, and now getting public-market capital).

The China robot angle (analyst)

Two shows put the U.S. build-out in global context. On The Negotiation, China analyst Rui Ma, who runs on-the-ground tech tours, said China has 200+ humanoid makers, of which only ~20 have raised $50M+, and named the leaders: Agibot, Unitree, and publicly-listed UB Tech. Real deployments are still narrow ("sorting and picking… moving things around… screwing things on") but concrete: Agibot ran a five-day livestream of robots doing quality assurance on tablets at Luxshare (the world's largest smartphone ODM). Her sharper insight: the near-term winner may be humanoid hands, not whole humanoids, one company is already making 10,000 robotic hands a month, most bolted onto simple arms rather than full bodies, because "a lot of the equipment is designed for human hands." And she flagged the coming trade wall: Americans can buy Chinese robots today, "but I don't know how much longer that will last" given the security concerns of a networked machine "sitting in your home." (Analyst voice.) The Negotiation

Sinica added the policy mechanism, and it's a genuinely useful frame. Law professors Angela Zhang and Alex Yang argued China's government now behaves like a "platform company", it doesn't own the champions (the fastest-growing EV, battery, solar, and humanoid firms are "predominantly privately owned" with small state stakes), it nurtures the ecosystem and sets the rules. Earlier this year Beijing convened industry, academics, and think tanks to write national standards for humanoid robots, which they argue is why Chinese robot costs "reduce so quickly every year", standardization drives supply-chain efficiency. The flip side is the recurring boom-bust: with 140+ humanoid companies (their count), massive overentry leads to brutal price wars and eventual consolidation, "this repeated pattern we have seen in battery and EV." The takeaway for Western investors: China's cost curve on robots is a policy weapon, and the same "profitless dominance" dynamic that hammered solar and battery margins is now setting up in humanoids. (Analyst/academic voice.) Sinica Podcast

4. Megaprojects: the binding constraint is power, not policy

If reshoring and AI are the destination, electricity is the road, and it's under construction everywhere at once.

The clearest picture of the power crunch came from Catalyst with Shayle Kann. The framing was sharp: to understand power today, "you don't have to ask about hyperscalers… you need to ask about AI labs", OpenAI and Anthropic are now the real demand drivers, with roughly half of Amazon's and Microsoft's new gigawatts flowing to them. The arithmetic doesn't work on the grid: data-center build-out is running at tens of gigawatts per year, growing ~50% annually with no sign of a slowdown, while the grid is adding only ~5-6 GW of gas and 20-25 GW (nameplate) of solar/battery per year. Anthropic alone wants to go from 1.5 GW of capacity at end-2025 to over 10 GW by 2027, "you're building a Google in two years." Since a gigawatt of capex runs $50-70 billion and a company "cannot make that kind of investment decision if there's uncertainty on the timeline," AI labs are going behind the meter, building their own generation. Concrete deals cited: OpenAI's 2.3-gigawatt reciprocating-engine project with Oracle in Shackleford County, Texas; XAI's aeroderivative turbines; Meta's behind-the-meter gas in Ohio; and Bloom Energy's fuel cells (now a ~$90B market cap) as the "ultimate play on power constraints." One host's important nuance: the harder constraint may not even be generation, it's transmission and distribution, especially high-voltage transformers that take three years to order. (Analyst/expert voices.) Catalyst with Shayle Kann

That transformer bottleneck is exactly where a great operator story sits. On the Powerline Podcast, executives from a company reshoring high-voltage transformer manufacturing (in partnership with utilities, at a Pittsburgh-area site) described the demand as "insatiable" and the opportunity as lasting "the next decade or more." Their frame on how much the world has changed: a 1-gigawatt data-center campus ("one nuclear plant pretty much full output," historically built to serve 800,000 residents) "is not so big anymore," with announcements now at 3, 5, and 10 gigawatts. They're building domestic capacity for 765 kV transformers, shunt reactors, and switchgear, with a major facility coming online end of 2027/early 2028, plus a smaller transformer plant in North Carolina aimed at data centers and renewables. The reshoring tell you can't get from a screen: they're hiring 300-500 workers at one site, starting around $27/hour, and building a full training complex, a real, wage-paying, blue-collar reshoring signal, straight from the people doing it. (Operator/insider voices.) Powerline Podcast

The political risk to all of this is now bipartisan and real. On Morgan Stanley's Thoughts on the Market, the public-policy team flagged that 75 data-center projects worth $130 billion were blocked or delayed in Q1 2026 alone, equal to the entire 2025 total (per Data Center Watch). Opposition is spreading from blue states (New York, Michigan, Illinois, Minnesota weighing moratoriums) into Pennsylvania, Arizona, Ohio, and parts of Texas restricting tax incentives. Their base case: a federal ban is unlikely under either party because the U.S. can't cede the strategic AI race to China, but expect a "conditional build-out" where projects must fund grid modernization and community benefits, and a growing share go fully off-grid (natural gas + fuel cells) precisely to sidestep water and air permits that communities weaponize to block them. A clean read-through: Bitcoin miners with existing grid interconnects are "clearly seeing a lot of demand" as a shortcut around the permitting wall. (Analyst voice.) Thoughts on the Market

The scale gap between the U.S. and everyone else is almost comic, and useful for framing where the industrial capital is actually going. On The Week with Roger, telecom analysts back from a German data-center conference described the disparity: all of Germany has ~3 GW of data-center capacity and expects ~1 GW of growth over ten years, while Elon Musk built 2 GW in under a year in Mississippi. U.S. data-center capex is now ~$750 billion, versus the "$70-90 billion" the telecom industry used to brag about; the Stargate project alone is 10 GW and expanding to two more campuses. Europe's problem is structural: a Northern Virginia approval takes ~3 months, the German equivalent ~7 years, and the Netherlands is "literally running out of land" and power. The supply-chain spillover they described is the real investable nugget: AI build-out has "bought out all the glass fiber for the next three years" (Corning), hard disks and memory are "sold out for the next three years," and spot DRAM prices rose 90% in three months. Anything upstream of a data center is being "gobbled up." (Analyst voices.) The Week with Roger

And the power theme has a clean investable menu. On Follow the Money, host Jerry Robinson pointed to the IEA projection that global data-center electricity use will more than double by 2030 to ~945 terawatt-hours, with U.S. power demand hitting record highs in 2026-2027, and named the beneficiaries he's watching: nuclear (Constellation Energy, NextEra) and uranium (Cameco). (Pundit voice, treat as thematic, not a specific call.) Follow the Money

Operators vs. pundits: how to weight this week

Weight the operators most. The highest-signal material this week came from people building the actual things: Peggy Johnson of Agility (real bookings, real unit economics, disciplined on valuation), the Path Robotics CEO (welding robots pulled by shipyards and data centers), the Powerline transformer executives (hiring 300-500 workers, naming voltage classes and timelines), the ChinaTalk gases veteran (the supply-chain vulnerability nobody prices), the Pur Gum founder (what a 39% tariff actually does to a P&L), and the Lumber Word mill traders (capacity permanently gone). These are the details you cannot get from a headline or a screen.

Weight the pundits for framing, not for calls. Dan Clifton (tariffs-as-stimulus, the Section 301 hand-off), the CSIS expert on USMCA, J.P. Morgan on copper, Morgan Stanley on data-center politics, and the Rui Ma/Sinica China-robot analysis are all valuable maps of the terrain, but they're commentary, not disclosure. Where a pundit named specific tickers (nuclear/uranium names on Follow the Money), treat it as a thematic starting point, not a recommendation.

The single most actionable synthesis: the constraint on the whole reshoring/AI/automation complex has moved from trade policy to physical infrastructure, power, transformers, permits, and 60 obscure gases. Tariffs are now a manageable, even stimulative, background variable; the megawatt and the transformer are the new chokepoints.

The week ahead: dates and catalysts to watch

  • Week of July 20: Third round of U.S.-Mexico bilateral trade talks. First real read on whether USMCA gets extended or fragments into bilateral deals. The Hard Country
  • Any day now: The overdue Section 232 copper tariff decision (Commerce report was due June 30). J.P. Morgan's base case is an escalatory outcome pushing copper toward $15,000/ton. At Any Rate
  • August (possibly October): Anti-dumping duties on Canadian softwood lumber drop from 35% to 25% (all-in ~45% to ~35%). The Lumber Word
  • Ongoing / new cadence: USMCA now faces an annual review, not a six-year one, so trade uncertainty is a permanent, recurring feature, not a one-time event. The Trade Guys
  • Filings to watch: Agility Robotics' S-4 (revenue detail beyond the $300M booked figure) as its SPAC merger progresses. StrictlyVC Download
  • Early 2027: First deliveries of Path Robotics' Rove welding-quadruped (50 units pre-sold, aimed at shipbuilding and data-center construction). The Robot Report
  • 2027 milestone: Anthropic's target of 10+ GW of power capacity, a proxy for how fast behind-the-meter generation (gas engines, fuel cells) must scale. Catalyst with Shayle Kann