When Iron Beats Code: The New Industrial Cycle

Sector Deep Dive

Market Trends

Defense

ISM Manufacturing New Orders hit 57.1 in January 2026 — a two-year high up 9.7 points in a month — while $660B in hyperscaler capex reveals a bottleneck: AI doesn't need more code, it needs 175 GW of electricity and physical infrastructure. Capital is rotating from virtual AI multipliers to iron.

HIGHLIGHTS

• ISM Manufacturing New Orders surged to 57.1 in January 2026 — a two-year high, up 9.7 points month-over-month — while the Industrial Momentum Indicator hit its highest level since December 2021, signaling a genuine manufacturing supercycle that hard data confirms and soft sentiment surveys have missed.

• Hyperscaler capital expenditure reached $660B in 2026 with focus shifting to energy and physical infrastructure — a 1-3% annual increase — but the true beneficiary is not the AI software stack: it is turbines, transformers, and the physical grid that must deliver power to data centers.

• US peak power demand forecasts rose +237% in a single year, from 38 GW to 128 GW, creating a 175 GW electricity deficit projected by 2033 — US power demand will exceed supply by 2028, making the grid the single biggest constraint on the entire AI industrial buildout.

• Siemens Energy ($ENR1N) holds a record €146B backlog with Q1 FY26 orders of €17.6B (+34% YoY, book-to-bill 1.82) — GE Vernova turbines are already sold out through 2028, large power transformer lead times exceed 2 years, and transformer prices are up +70% since 2019.

• 55% of the US grid is more than 33 years old — a structural replacement cycle that would be mandatory even without the AI demand shock, creating a decade-long visibility window for grid infrastructure manufacturers entirely independent of any single technology cycle.

• $CSCO (Cisco) reported FQ2 2026 revenue of $15.3B (+10% YoY) with AI orders jumping +62% QoQ to $2.1B and an FY26 AI order target raised to >$5B — trading at ~18x forward P/E versus triple-digit multiples for AI software, representing one of the widest valuation discounts in the sector.

• Boeing ($BA) logged its best order month since 2012 in January 2026 with a backlog of ~6,700 aircraft — Delta Air Lines ordered 30 x 787-10s, breaking Airbus's near-monopoly — while NATO's 3.5% GDP defense target and EU SAFE initiative provide a guaranteed cash flow backstop through 2030.

• Value-factor positioning is at a +3 standard deviation relative to growth's -3 sigma reading — the factor rotation is statistically extreme and historically persistent once initiated, consistent with the 2003-2004 post-DotCom rotation and the early 1990s manufacturing recovery cycles.

• NACM manufacturing credit data signals 3.5%+ US GDP growth for 2026, hard data leads EPS by approximately 13 months, suggesting the current industrial earnings acceleration is still in its first quarter — the majority of EPS catch-up has yet to be recognized in analyst estimates.

• Portfolio construction: overweight real assets and large-cap industrials with backlog visibility ($ENR1N, $BA); underweight pure software with no FCF and high multiples; pair trade long Siemens Energy / short AI-software basket; hedge with gold ($2,800-3,200 target) and short-duration bonds; execute in phased entry over 3-6 months with max position size 20-25%.

‍

EXECUTIVE SUMMARY

A regime change is underway in global capital allocation. For the past two years, markets have rewarded virtual AI multipliers — software companies trading at triple-digit multiples on the promise of intelligence at zero marginal cost — while systematically undervaluing the physical infrastructure required to actually run AI at scale. The signal that this regime is ending arrived in January 2026: ISM Manufacturing New Orders surged to 57.1, a two-year high up 9.7 points month-over-month, while the Industrial Momentum Indicator reached its highest level since December 2021. The factor rotation has turned statistically extreme, with Value at +3 standard deviations relative to Growth's -3 sigma reading. This is not a correction in AI stocks; it is a durable rotation toward companies that make the physical world function.

The macro architecture supporting this thesis is unusually robust. NACM manufacturing credit data signals 3.5%+ US GDP growth for 2026, and hard data leads corporate EPS by approximately 13 months — meaning the majority of the industrial earnings acceleration is still ahead of what analyst consensus has priced. The Fluid Power Survey reads 73 (expansion), the Truck Demand Index hit its maximum since April 2022, and semiconductors are growing +30% year-over-year with memory up 48%. Manager surveys still cite recession risk and uncertainty, creating the classic setup where sentiment lags reality — the precise condition that characterized both the 2003-2004 post-DotCom rotation and the early-1990s manufacturing recovery, historical parallels the report identifies explicitly.

The central thesis is that AI's true bottleneck is electricity, not code. Hyperscaler capital expenditure reached $660B in 2026, focused on energy and physical infrastructure — yet US peak power demand forecasts rose +237% in a single year, from 38 GW to 128 GW, creating a projected 175 GW electricity deficit by 2033. US power demand will exceed supply by 2028. The December 2025 PJM capacity auction saw prices rise by $6.5B due to data center buying, validating the scarcity premium forming in physical power markets. GE Vernova turbines are sold out through 2028 and are already booked at ~$2,500/kW; large power transformer lead times exceed two years; and 55% of the US grid is over 33 years old — a replacement cycle mandatory regardless of AI demand. The physical world does not scale like software.

Three companies frame the tactical playbook. Siemens Energy ($ENR1N) carries a record €146B backlog with Q1 FY26 orders of €17.6B (+34% YoY) and a book-to-bill ratio of 1.82; the 2028E P/E drops to ~21x as the earnings ramp executes, with a base case target of €170-180 and a bull case of €200+. Cisco ($CSCO) reported FQ2 2026 revenue of $15.3B (+10% YoY) with AI orders +62% QoQ to $2.1B and FY26 target raised to >$5B, trading at just ~18x forward P/E versus triple-digit software multiples — a cash machine with $13.3B FCF at a structural discount. Boeing ($BA) logged its best order month since 2012 with a ~6,700 aircraft backlog and a projected FCF hockey stick from $1.2B in 2026 to $14.5B by 2028.

Risk management requires discipline across three identified gray rhinos. Physical grid failure — the scenario where data centers cannot secure power, forcing hyperscalers to cut capex and crashing the entire AI value chain — is the Energy Stop tail risk. CLO ETFs (+58%) and private credit instruments that have never been stress-tested represent a private credit bubble with hidden leverage that could cascade across industrial credit in a recession. Geopolitical re-armament (NATO 3.5%) structurally embeds inflation into the long end of the rate curve, keeping long-duration borrowing costs elevated and compressing multiples for capital-light technology. The portfolio response: maintain overweights in large-cap industrials with confirmed backlogs, hedge via gold and short-duration bonds, and execute the Siemens Energy / AI-software pair trade to directly capture the rotation from virtual to material capital allocation.

‍

Text Link