1. AI Compute, Hardware, and Data Centre Capex: Record-Scale Deals and Power-Grid Stress
Description:
The week of April 20–27 produced the largest single compute supply contract in AI history, alongside continued evidence of a hyperscale infrastructure arms race. Combined capex guidance from the top five AI/cloud providers now approaches $690B for 2026 (nearly double the prior year) while energy, land, and component constraints signal that access for non-contracted buyers will tighten significantly through 2028.
Key Signals:
CoreWeave–Meta $21B Compute Supply Deal (announced ~April 20, 2026): The largest multi-year cloud GPU infrastructure contract to date, running through 2032. Sets a new benchmark for long-cycle compute procurement and signals how the largest AI spenders are securing capacity well in advance.
CoreWeave–Anthropic Multi-Year Agreement: A further multi-year compute agreement cited as exemplary of non-Big Tech cloud alternatives gaining traction; detailed terms not disclosed.
Nebius Hyperscale Expansion: Scaling from 170 MW (2025) to approximately 1 GW by end-2026, with $16–$20B in 2026 capex committed — a notable non-US/European infrastructure expansion signal with implications for regional compute access.
Top Five Global AI/Cloud Providers — $660–690B Combined 2026 Capex: Nearly double the prior year; reflective of a sustained infrastructure arms race that is reshaping barriers to AI entry across industries and geographies.
WEF $7T AI Buildout Context: Global AI infrastructure investment projections have reached the $7T range, providing broader context for the scale and duration of the current capex cycle.
Potential Impact:
Sustained hyperscale investment is likely to reinforce US and allied dominance in AI compute, with smaller independent clouds gaining share as hyperscalers diversify supply.
Energy, land, and GPU/accelerator component constraints will raise barriers to entry and create regional bottlenecks — particularly for organisations without existing supply agreements.
Strategic buyers able to secure multi-year contracts now will gain durable cost and performance advantages as capacity tightens into 2028.
Stage of Adoption:
Transactions closed; infrastructure build-out in progress. Market-wide pricing and access effects will be felt throughout 2026 and into 2027–28.
Implication:
AI access, working capital, and cost structures for enterprises and startups will depend heavily on hyperscaler relationships and their supply prioritisation. Compute scarcity risk is real and approaching for non-contracted buyers.
Sources:
2. Regulation, Policy, and Sovereign AI: Compliance Deadlines Loom, Quiet Period for New Announcements
Description:
No major regulatory, policy, or sovereign-AI announcements were issued by official agencies, ministries, central banks, or standards bodies during April 20–27, 2026 across APAC, LATAM, GCC, Europe, or North America. The enforcement calendar, however, is not quiet as a series of staggered deadlines across the EU, Korea, and Colorado are approaching, and SME and hyperscaler exposure to "high-risk" classifications is growing.
Key Signals:
EU AI Act — August 2026 Phase-In: Amendments from March 2026 introduced application delays for high-risk AI systems (next major phase August 2026, then December 2027) and new bans on nudifier apps, alongside increased demands for transparency and SME support measures.
South Korea — Framework Act (January 2026): Established a National AI Committee, oversight requirements for high-impact AI systems, and mandatory labelling for generative AI outputs — with enforcement timelines now approaching.
GCC — Sovereign AI Stack Development: Ongoing strengthening of data residency, ethical AI standards, and localised infrastructure in UAE and KSA; sovereign AI investment continues to outpace Western approaches in speed-to-deploy.
LATAM — Google/IDB Policy Partnerships (April 2026): Regional calls for stable frameworks in Brazil and Colombia tied to data centre and energy expansion, with Google partnering on three new AI policy and innovation initiatives across the region.
North America: No new central bank, executive order, or standards bulletin signal during the reporting period.
Potential Impact:
Monitoring now required for August 2026 EU, Korea, and Colorado deadlines; SME and hyperscaler exposure to "high-risk" regulatory burdens grows with each phase-in.
Compliance planning must anticipate staggered but accumulating requirements across jurisdictions; local adaptations are increasingly essential, especially outside the US/EU core.
Stage of Adoption:
In-force or imminent in most major markets; phase-in periods for application and enforcement vary by jurisdiction. No immediate new regulatory requirements, but enforcement thresholds are approaching.
Implication:
Short-term regulatory exposure is steady, but from Q3 2026 enforcement risk increases sharply in high-risk sectors. Non-compliance may result in market exclusion or penalties across EU and Korea. US/EU divergence persists as a long-term strategic risk.
Sources:
3. Patent and IP: Favors Human-in-the-Loop, Not AI-as-Inventor
Description:
The operative landscape is shaped by a cluster of recent-but-prior signals that collectively define a tightening framework: human inventorship is mandatory, technical improvement must be explicit, and disclosure requirements are rising globally.
Key Signals:
USPTO Ex parte Desjardins (Dec 2025): Reversal of a §101 rejection for AI claims addressing "catastrophic forgetting" validated technical improvements as patent-eligible, triggering updated examiner guidance that now favours well-documented AI innovations and lowers rejection rates for genuinely novel technical contributions.
USPTO Revised Inventorship Guidance (2025): Human inventorship is required even when AI is involved in the creative process. AI assistance does not negate individual inventorship provided conception criteria are met — a clarification that expands the scope of protectable AI-assisted work.
EPO Board of Appeal T 1669/21: Heightened disclosure requirements for AI/ML patents now require patentees to provide explicit detail on model design and training methodology. Vague or high-level descriptions will not meet the bar.
EPO / UKIPO — DABUS Ruling Upheld: Continued rejections of AI-as-inventor claims reaffirm the "natural person" inventorship standard across European jurisdictions.
UK Supreme Court, Emotional Perception v. Comptroller (Feb 2026): Overturned the outdated UK test for AI/software patentability, aligning with EPO's "technical effect" doctrine and broadening eligible subject matter for AI-related filings in the UK.
WIPO: Ongoing facilitation of global IP harmonisation and development of AI-enabled IP tools, with convergence toward the technical-effect standard gaining traction across member states.
Potential Impact:
Patent offices are reducing barriers for algorithmic and domain-specific inventions (ML, life sciences) where technical improvement is clearly articulated but human inventorship and rigorous enablement remain non-negotiable.
Filings relying on AI as sole inventor, or lacking a clearly demonstrable technical contribution, will continue to be rejected across all major jurisdictions.
Enterprises that can demonstrate human-led inventive steps and detailed technical disclosure stand to significantly expand their AI patent portfolios.
Stage of Adoption:
Examiner guidance and prosecution practice are actively evolving; international convergence toward the technical-effect and human-inventorship standard is growing but not yet fully harmonised.
Implication:
AI patenting accelerates for well-documented, human-originated innovations. Only organisations that can meet rising technical disclosure standards will capture the IP advantage. The bar is moving up, not down.
Sources:
4. Enterprise and Sector AI Adoption: Revenue Mix Shifts and Measurable Productivity Claims
Description:
AI has crossed the threshold from pilot curiosity to revenue contributor at leading professional services and enterprise firms. BCG reported that 25% of its 2025 consulting revenue was AI-related, while McKinsey survey data confirms 88% of global enterprises now use AI in at least one function. The lead-lag gap between AI-mature and AI-lagging organisations is beginning to surface in measurable EBIT and productivity outcomes.
Key Signals:
BCG 25% AI Revenue Mix (April 23, 2026): $3.6B of BCG's $14.4B 2025 consulting revenue was AI-derived representing a structural shift in professional services revenue mix and signalling where enterprise spend is flowing.
McKinsey Global Survey (2025): 88% of global enterprises use AI in at least one function; 65% deploy generative AI; 62% are piloting AI agents; nearly 40% report quantifiable EBIT impact. The majority, however, remain in pilot or prototype stages rather than scaled deployment.
NVIDIA State of AI Report (2026): 53% of surveyed sectors report productivity increases attributable to AI; 34% report new revenue streams. Marquee deployments (including PepsiCo) are delivering over 20% throughput improvements, near-100% design validation rates, and 10–15% capex savings.
Potential Impact:
AI's direct effect on margin, cost, and productivity is increasingly observable at scale — particularly for multinationals leveraging digital twins and supply-chain AI.
Professional and IT services spend continues to shift toward AI-related work, putting pressure on traditional software and services margins.
Sector leaders with successful scaled deployments have begun capturing measurable EBIT and revenue uplift versus laggards.
Stage of Adoption:
Rapid expansion from pilots to production in telecom, CPG, financial services, and manufacturing; a pronounced lead-lag dynamic is emerging between enterprise AI leaders and followers.
Implication:
AI-driven investment now represents a quarter of global consulting revenue. Enterprises lagging in scaled adoption risk measurable margin erosion and competitive disadvantage by FY28.
Sources:
