Physical AI Projected to Grow 20× by 2036 Nelson Tsay, president of Ennoconn ( a Taiwanese industrial computing company within the Foxconn Group, expects physical AI to take off in 2026 Tsay cited research projecting the physical AI market to grow from $6 billion in 2026 to $120 billion by 2036, a 35% CAGR. Humanoid robot shipments are expected to exceed 50,000 units in 2026, up more than 700% YoY Ennoconn operates three US factories with more than eight SMT lines, over 40 SMT lines in Europe, and manufacturing facilities in Southeast Asia

Physical AI prototypes are everywhere on trade show floors. Ambarella measures readiness by a different standard: fleets running reliably for years, upgraded and monitored across thousands of deployed units. At AI Infra 2026, @BrendanBurkeX and @MattKimball_MIS talk with Muneyb Minhazuddin, Customer Growth Officer at @Ambarella_Inc, about a 22-year chip architecture cutting memory use 10x. $AMBA

Physical oil prices in the North Sea and Mediterranean have surged as some refiners seek to replace prompt Saudi barrels loss after the East-West pipeline was hit. Forties crude is now trading +$15 a barrel over Dated Brent, compared to $0.5 a barrel in early September.

“Physical AI” is having a moment. At Zoox, it’s been what we’ve been building since 2014. Hear Co-Founder and CTO Jesse Levinson delve into where we started, where we're at, and where we're going:

Physical AI's real bottleneck isn't the model. It's the data. 📊 @Voxel51 + Nebius built a closed-loop pipeline to fix it: - Curate in FiftyOne - Generate at scale on Nebius - Review in FiftyOne Rare edge cases → solved.

Physical AI is heating up ▶️ Global humanoid robot shipments exceeded 22,000 units in 1H26, up 300% YoY ▶️ Counterpoint expects more than 50,000 units in 2026, representing about 210% YoY growth China dominates the market AgiBot shipped about 9,700 units for a 43.1% share, while Unitree shipped more than 7,000 units for a 31.1% share Together, they accounted for about 75% of global shipments, while the top five vendors represented 86% By use: · Entertainment & performance: 33.6% · Data production & research: 27.0% · Service & guidance: 19.0% · Intelligent manufacturing: 12.8% · Warehousing & logistics: 4.9% The main shift is toward real industrial deployment AgiBot’s G2 robots are being introduced into production environments through partnerships with Longcheer Technology and Joyson Electronics, while Galbot has tested its S1 industrial robot on CATL production lines. Leju’s Kuavo robots are being used in robot-data production and training centers UBTECH raised its 2026 Walker S shipment guidance from 2,000–3,000 units to 5,000, citing stronger industrial demand Counterpoint expects more service and industrial pilot projects to transition into mass deployment during 2H26 Competition is increasingly shifting from simply producing capable hardware toward complete deployment ecosystems combining advanced models, vertical applications, data systems, and rapid model iteration

Physical AI becomes yet another growth vector for $AMD. Kodiak is putting AMD EPYC processors at the heart of its seventh-generation driverless truck platform — the first company to deploy them in driverless trucking hardware. Higher performance. Lower power. Faster real-time processing of the massive sensor data these trucks need to operate safely with no one in the cab. Already running fully driverless in the Permian. Targeting public highways later this year. As autonomy scales, the compute powering it carries increasing profit potential.

Physical AI companies are about to learn an old infrastructure lesson: A customer contract can be or unlock a form of capital. Gatik just announced a $200 million raise, but the more important numbers are the ones sitting behind the round: more than $600 million in contracted revenue, 85,000 fully driverless orders completed and 99% on-time delivery, all reported by the company. The contract details are not public and neither are recognized revenue, margins or capital required per deployed lane, all of which matter. However, investors across both equity and debt can look at backlog in two buckets (i) high-quality demand or (i) a potentially expensive promise. Nonetheless, I think this is still a useful signal/lesson for Physical AI founders. First, contracts can de-risk demand before a company scales the fleet. A signed route or workflow is more financeable than a general claim that a machine has a large market. Second, operating proof can separate technology capital from asset capital. Equity will fund building out autonomy, safety and the next product. Once the deployed unit is predictable, equipment finance or other lower-cost capital can fund more trucks or robots. Third, contracts force repeatability. A customer buying reliable throughput cares about on-time delivery, interventions, uptime and cost per order. After things are live in the field, your ability to perform is whats up for the test. Lower cost debt capital is a fundamental need for the current Physical AI valuations to fundamentally work longer-term if you are handling hardware. Therefore, being extremely intentional during contracting is critical. The top 5 things that would be on my mind as a Founder are: 1. Duration of contract 2. Cancellation rights 3. Minimum volume 4. Your own deployment capex and payment terms with suppliers 5. Your expected gross profit and what levers you need to believe in for it to be X vs. Y If you can pull forward customer payments to finance the business that's even better. But if you can't, making sure your contracts and ability to back it up are airtight is key.