GlobalData forecasts the total AI market will be worth $642bn in 2029, up from $131bn in 2024, pointing to years of breakneck expansion ahead. Yet the most important change is qualitative, not just financial. The center of gravity is shifting from systems that draft and summarize to systems that operate. “Agentic” tools – software that can plan tasks and take action with limited human supervision – are moving from demos into product catalogs. Many buyers still hesitate over whether such tools can produce measurable business value, but that hesitation itself is revealing. It suggests AI is becoming rooted in everyday commercial life even before the return on investment is fully agreed, which is exactly how big platform shifts often arrive.
The new battleground sits beneath the model
Beneath the headlines touting AI’s unstoppable advance, battles are being waged to determine who will dominate an AI-driven future. Infrastructure is one of the most decisive fronts. Nvidia still dominates AI chips, with just under 90% market share by revenue. But the market is starting to broaden, and buyers are starting to behave like buyers rather than supplicants.
Compute is becoming a procurement problem, not a technical one
OpenAI, previously reliant on Nvidia GPUs, has diversified with multibillion-dollar commitments – 10 gigawatts of custom accelerators from Broadcom and 6GW of GPU capacity from AMD – while Qualcomm is expected to launch inference chips in 2026. This is less a sudden upset than the opening of a long campaign. As more credible alternatives appear, large purchasers will hedge supply and use competition to push prices down. In an industry built on scarcity, that is a structural shift, not a footnote.
Geopolitics is reshaping supply chains
Geopolitics is pushing in the same direction. Question marks loom over Chinese dominance of chip production, with export controls and industrial policy reshaping who can sell what, and where. For some of today’s AI leaders, exposure to that tension is turning into a risk factor rather than a growth lever. Nvidia reported revenue from China, which has historically contributed as much as one-fifth of all revenues to the company, falling to zero in October 2025.
The physical limits: power, water and cooling
Then there are the physical constraints, the sort that do not respond to a clever press release. A single training run for a frontier model can demand dozens of megawatt-hours. Cooling adds another pinch point, with high-density AI workloads consuming millions of gallons of water per data center per year. Availability of power and water – or lack of it – could yet hem in the boom. The AI economy may look like software, but it increasingly behaves like heavy industry.
Across the business world, leaders therefore face a double task. They need to identify AI’s productivity potential, and quickly, while staying alert to the risk of a bubble whose bursting would have repercussions far beyond the offices of Nvidia and OpenAI. What should their roadmap for 2026 look like?
Key themes in AI
The coming year’s key stories will be shaped by a simple reality: systems are becoming more capable and more independent, and enterprises are being asked to trust them with real work.
Agents: autonomy with guardrails
One theme will be the challenge of harnessing increasingly autonomous systems without surrendering control. Agentic AI promises to compress long chains of routine work – extracting information, filing forms, scheduling tasks, reconciling records – into shorter cycles. New models, like Anthropic’s Claude Cowork, can do all this and more at the touch of a button. The attraction is obvious, but so is the trap. Autonomy in business operations is essential, yet ungoverned autonomy is a liability. In 2026, the enterprises that move from experimentation to durable value will be those that weave autonomy into core operations with intent: finance, customer support triage, IT service desks. Where decisions carry regulatory or safety risk, firms will demand systems that can explain actions, log each step and hand off cleanly to humans. Leaving everything to AI remains too risky an option.
Sovereignty: where the model runs matters
A second theme is sovereignty. As models become embedded in core processes, questions about where data sits and who controls it will keep C-suite members up at night. AI products are spinning off market-leading LLMs to preserve data control while deploying AI across hybrid environments tied to specific countries or regions. This matters most in organizations that must reconcile cloud scale with jurisdictional rules, and it is becoming a board-level question rather than an IT preference. In 2026, “where does the model run?” will matter nearly as much as “what can it do?” In the face of rising regulatory scrutiny, inarticulacy is not an option.
Capital and consolidation
The maturing AI capital cycle is another trend worth watching. Cashflow into AI shows no sign of slowing, but it may become more selective as investors and buyers learn which parts of the stack compound and which merely glitter. In late 2025, after Nvidia reported Q3 2025 revenue of $57bn, up 62% year on year, even optimists were surprised by the market’s continued momentum. Major purchases are continuing apace too, including Meta’s 49% purchase of Scale AI for $14.8bn. The pattern is of an industry consolidating around scarce inputs: data center capacity, specialized chips, and the tools that make models dependable in production.
Scarcity in the real world
Finally, there is infrastructure scarcity itself, the hard floor beneath the soft talk. GlobalData projects the cloud computing market will be worth over $1tr in 2026. The AI chip market is forecast to grow rapidly through 2030, with GlobalData charting a 51% CAGR for 2022–2030. Yet supply constraints, tariffs and local opposition to new data centers are raising costs and stretching timelines. Some providers may opt to delay or scale back expansions because utilities cannot add capacity quickly enough. The market’s expansion is dramatic, but not inevitable. Businesses integrating AI tools at speed should plan for the possibility that physical restrictions impede scale, and that today’s “just add compute” assumptions fail at the worst moment.
Industry-level insights
If AI is becoming a general-purpose technology, it is still being applied in highly specific ways. Each sector will meet different opportunities and different risks, which is why investors and dealmakers should resist the temptation to treat “AI exposure” as a single valuation multiple.
Consumer: personalization gives way to operational advantage
Start with the consumer sector, where the early wins have often been presented as marketing magic. Personalization – product recommendations, targeted promotions and service that adapts to context – is likely to overhaul business as usual in consumer goods, foodservice and hospitality. Yet the next wave is operational rather than promotional. Agentic systems that restock inventory, adjust schedules and respond to demand swings are being rolled out precisely because margins are thin and decisions are repetitive. GlobalData analysis found AI-driven forecasting and supply planning at Unilever reduced inventory levels by about 16%. That is the kind of outcome investors should watch: fewer pallets, less waste, more efficiency. AI here is not better ad copy. It is a different operating rhythm.
Healthcare and life sciences: adoption through scrutiny
In more technical sectors, the numbers are both clearer and larger. GlobalData values the combined AI market across pharmaceuticals and medical devices at $11.9bn in 2024, reaching $57.4bn by 2029, implying growth of roughly 37% a year. Drug discovery, imaging and operational automation all represent major use cases, but an additional accelerant is policy alignment. Regulatory approvals and reimbursement programs are giving hospitals confidence to deploy trailblazing imaging tools and integrate them into clinical routines. Over the next twelve months, pharma, healthcare and other science-forward sectors can use AI not only to push the frontier, but to prove safety, meet documentation requirements and slot into workflows without slowing clinicians down. The prize is not disruption for its own sake, but adoption that survives scrutiny.
Industrial sectors: steady gains, slower transformation
AI is still finding its feet in heavy industries, though even there the outlines are sharpening. GlobalData projects construction AI spending will exceed $25bn in 2029, well below the $41.3bn expected in manufacturing. But use cases are becoming clearer. Generative AI tools in construction could be worth $5bn by 2029, driven by tools that catalyze design, documentation and information search. Faster inspections, better safety monitoring and fewer errors in measurement are nearer-term opportunities, and they are easier to justify than grand claims about automated megaprojects. As with pharmaceuticals and healthcare, compliance requirements can be met and even anticipated with AI. Sustainability reporting is a case in point. Proactive AI-driven design tools can expose trade-offs early and help mold greener manufacturing choices, turning regulation from a drag into a design constraint that is managed, not merely endured.
What this means for investors and deal teams
Across these industries, the managerial takeaway for 2026 is consistent. Models are improving and agents are proliferating, but the firms that do best will treat AI as an operational catalyst – measured, auditable and tied to specific decisions – rather than as a magic wand to be waved over a spreadsheet. For investors, M&A dealmakers and corporate strategists, that framing matters. It changes due diligence questions, shifts integration priorities and clarifies where defensible advantage can be built.
Discover further insights
To learn more, download The Future of Tech, 2025–2035: Insights for Investors & Dealmakers, published in association with Sterling Technology – the provider of premium virtual data room solutions for secure sharing of content and collaboration for the investment banking, private equity, corporate development, capital markets and legal communities engaged in technology and TMT M&A dealmaking and capital raising.
