6 mins read
Editor's Note
This week's five stories share a single, uncomfortable finding: AI is already reshaping the economy — and the costs are landing before the productivity gains.
Goldman Sachs, JPMorgan and Stifel now agree the AI infrastructure boom is measurably inflationary today, driving up electronics prices even as firms promise future efficiency dividends. The layoff data makes the same point from a different angle: over 92,000 tech workers have lost jobs in 2026 so far, while the four largest tech firms redirect $725 billion into compute rather than payroll. Robotics is the one sector where the productivity thesis is already delivering — the global market grew 34% this year, with AI-native systems displacing older automation models. And the Five Eyes security warning closes the week with a structural caution: organisations deploying autonomous AI agents inside critical infrastructure are doing so faster than anyone has built a governance framework to contain them.
01
The AI Buildout Is Inflationary Now: Goldman, JPMorgan and Stifel All Agree
Goldman Sachs, J.P. Morgan Asset Management and Stifel published research this week finding that the AI infrastructure boom is creating measurable inflationary pressure today. Surging demand for AI infrastructure has driven up prices for digital memory and storage batteries, rippling across the consumer electronics supply chain. Goldman's equity analysts forecast average selling prices for computers and non-Apple smartphones will rise by roughly 10% this year. J.P. Morgan Asset Management's chief global strategist noted that memory chip prices have soared due to AI buildout demands, raising costs for manufacturers of laptops, smartphones and automobiles. Stifel's Thomas Carroll called this a macro "regime shift" — the first time in 65 years that tech goods prices are rising faster than wages.
Why it matters: Finance teams modelling AI's productivity dividend need to account for its concurrent inflation cost. Companies absorbing higher electronics and infrastructure pricing now — while waiting for AI efficiency gains to materialise — face a compressing margin window. The firms best positioned are those that locked in procurement and capex before the current pricing cycle.
Source: Fortune, 4 May 2026
02
92,000 Tech Workers Cut in 2026 — AI Now Named as Primary Structural Driver
More than 92,000 tech employees have been laid off across 95+ companies so far in 2026, according to Layoffs.fyi data cited by CNBC — an average of over 882 job losses per day. Amazon has cut roughly 30,000 roles since October; Oracle eliminated up to 30,000, approximately 18% of its workforce, to fund $156 billion in AI infrastructure. Meta confirmed cuts of 8,000 employees beginning 20 May, explicitly linking the reduction to its $115–135 billion AI capex budget. What distinguishes this wave from previous tech cycles: companies reporting record revenues are simultaneously cutting headcount and redirecting the savings into compute. Meta's Q1 2026 revenue reached $201 billion, up 22% year-on-year; free cash flow hit $43.6 billion.
Why it matters: This is no longer a correction — it is a structural shift in how large technology companies allocate capital. For HR and finance leaders across all sectors, the pattern signals that AI-driven headcount reduction is not a cost-cutting measure but a reallocation strategy. Boards should be asking: at what point does the same logic apply beyond the technology industry?
Source: CNBC, 24 April 2026
03
Global Robotics Market Hits $38 Billion, 34% increase from previous year
The global robotics market reached $38 billion in 2026, up 34% year-on-year — the fastest growth rate in a decade — according to the Silicon Valley Robotics Center's annual State of Robotics report. Vision-Language-Action (VLA) models, absent from production deployment 18 months ago, now underpin 40% of all new commercial deployments. Logistics and warehousing account for 41,000 deployed units; semiconductor manufacturing for 22,500; food service for 8,200 units — a 61% year-on-year rise. The cost of teleoperation data collection has fallen 65%, from $340 per hour in early 2024 to $118 per hour as of March 2026, putting enterprise pilots within financial reach for the first time.
Why it matters: Robotics is the one sector where the AI productivity thesis is already delivering at scale. For industrial, logistics and food service operators, the data cost reduction is the decisive signal: the $50,000–$150,000 pilot budget that once required board-level approval is now a standard line item. Companies not actively piloting face a compounding disadvantage as competitors accumulate proprietary training data.
04
Big Tech's $725 Billion AI Capex Now Exceeds Global Oil and Gas Investment
Amazon, Microsoft, Alphabet and Meta collectively plan $725 billion in capital expenditure in 2026 — a 77% year-on-year increase — directed almost entirely at AI data centres, custom chips and foundation models. The International Energy Agency confirmed separately that the capex of just five major technology companies now exceeds global investment in oil and gas production. Meta's CEO linked the company's 8,000 announced job cuts directly to this budget reallocation. Amazon's Q1 2026 operating margin reached 13.1% — its highest ever; Alphabet's reached 36.1%; Meta's hit 41.4%. These four companies are spending unprecedented sums while producing record operating margins — a combination not previously seen at this scale in corporate history.
Why it matters: When compute investment surpasses global oil and gas capex, energy and infrastructure assumptions across industries need revisiting. For investors, the concentration risk is acute: four companies now account for a structural portion of global infrastructure spend. For every other sector, the critical question is whether AI investment of this magnitude will generate the productivity gains needed to justify the price inflation it is simultaneously creating.
Source: 24/7 Wall St., 7 May 2026; IEA, April 2026
05
Five Eyes Issue First Joint Alarm on Agentic AI: "Organisations Must Assume It Will Behave Unexpectedly"
On 1 May 2026, cybersecurity agencies from the United States, United Kingdom, Australia, Canada and New Zealand published their first joint guidance on agentic AI — systems that autonomously plan, decide and take actions across enterprise networks. CISA, the NSA and their Five Eyes counterparts warned that autonomous agents are already operating inside critical infrastructure and defence sectors, with most organisations granting them far more system access than can be safely monitored. The document identifies five risk categories: privilege escalation, design flaws, unpredictable behaviour, structural cascade failures, and accountability gaps. It states that organisations should "assume that agentic AI systems may behave unexpectedly" until security practices mature, and recommends staged rollouts beginning with low-risk tasks only.
Why it matters: Five Eyes agencies do not publish joint guidance on emerging technology lightly. For CISOs and boards, this document is now the baseline for agentic AI governance — and the gap between what it recommends and what most enterprises currently have in place is significant. Any organisation that has deployed AI agents in customer-facing, financial or infrastructure systems without a formal privilege and audit framework should treat this as an urgent gap to close.
This Week's AI Tip
When you use AI for analysis or complex decisions, the default response is a conclusion. The problem: the model may have skipped important steps, made a hidden assumption, or reached a plausible-sounding answer without genuinely working through the problem. Adding four words — "think through this step by step" — changes the output fundamentally.
This technique, called chain-of-thought prompting, forces the model to reason sequentially before committing to a conclusion. It dramatically reduces errors on strategic, analytical or multi-variable tasks — the exact kind professionals face daily.
The before/after below illustrates the difference for a real strategic decision:
Before
"Should we expand into the German market?"
After — try this prompt
"Think through this step by step: we have a £500k budget, a B2B SaaS product, and no existing EU presence. What are the key factors for and against expanding into the German market this year, and what would need to be true for this to be the right move?"
🌿 Good News · AI Making a Difference
AI Energy Efficiency Per Task Is Improving at a Rate "Unprecedented in Energy History"
The International Energy Agency's latest report on energy and AI contains a finding that cuts against the dominant narrative of AI as an insatiable energy consumer. Software and hardware advances have reduced the electricity consumed per individual AI task by at least an order of magnitude annually in recent years — a pace the IEA describes as unprecedented in energy history. Simple text queries now consume less electricity than running a television over the same period. The caveat is real: more people are using AI and energy-intensive use cases such as AI agents are rising, so total consumption continues to grow. But the efficiency trajectory means that AI's per-unit energy footprint could decouple from usage growth if innovation continues — a meaningful signal for organisations managing emissions commitments alongside AI adoption.
If you’ve been forwarded this email, you can subscribe here
We’d love to hear your feedback or any suggestions: email us at [email protected]
