AI Leaders Are Pulling Ahead
AI leaders are pulling ahead
A report worth reading carefully
Every year, Bain & Company publishes its Technology Report. This year's (2025) title is: "AI leaders are extending their edge."
The report was written by the team of David Crawford, chairman of Bain's global TMT practice, based on two years of tracking the AI wave and on empirical company data. Unlike the "trend sketches" of years past, this time Bain seems to be opening a file on a brand-new world order: AI is becoming the central variable that determines wealth and the competitive landscape between nations.
For us — whether investors, entrepreneurs or managers of family offices — this is not just a technology report but a map of where capital will flow over the next decade.
The AI leaders' flywheel: winners in an age of compounding
Bain points out that 2025 has become the inflection point of AI divergence.
Among the world's twenty most valuable tech companies, the top five (Microsoft, Amazon, Alphabet, Meta, Nvidia) account for more than 70% of total market value.
Nvidia's market cap alone has grown more than 800% since 2023.
These giants are forming a "self-compounding" closed loop —
they no longer just do AI; they use AI to reshape the three-layer ecosystem of infrastructure, models and applications.
Meanwhile, a crop of nimble challengers is breaking through at every layer:
Foundation layer: start-ups such as CoreWeave, Mistral and Anthropic are tearing open entry points in GPU cloud and model innovation;
Application layer: Anysphere's AI code editor Cursor and Perplexity's AI browser are redefining the boundaries of "human-machine collaboration".
AI is no longer a product but a systemic self-evolution.
Sovereign tech and a fragmented world
Bain defines this era as "Sovereign Tech, Fragmented World".
It is a global realignment around "who controls the lifelines of technology".
The US leads in chips and large models, reshaping supply chains through export controls and tariffs;
China is accelerating semiconductor self-sufficiency, already holds about 20% of global wafer capacity, and is breaking through in models and robotics;
Europe has launched the €200 billion InvestAI plan to build homegrown AI gigafactories;
New players in the Middle East, such as Saudi Arabia's Humain, are building 500-megawatt-class AI data centres and developing Arabic multimodal large models.
The "sovereignisation" of AI is pushing the world towards multipolarity:
AI systems will increasingly resemble national products, reflecting their own political and cultural contexts.
This means tomorrow's global companies will need to understand not only market logic but also "algorithmic geopolitics" — finding ways to stay compatible and survive across different sovereign AI systems.
Capital logic reshaped: the end of software's "easy growth" era
The report notes that tech M&A accounted for 22% of all deals in the first half of 2025, but "the era of easy growth in software is over".
Over the past decade, software investing generated returns through scale and multiple expansion;
over the next decade, value will come more from operational efficiency and AI-enabled innovation.
Bain's data reveals:
since 2010, 53% of investment returns have come from revenue growth, while the contribution of multiple expansion is declining;
investors will have to rely on more refined forms of growth — AI-driven product iteration, data monetisation, cross-regional expansion.
This is also an era that shifts from "riding the wave" to "sculpting growth".
Agentic AI: the next paradigm after SaaS
In Bain's view, the next act of AI is called Agentic AI.
It means software begins to have "the power to execute" — to reason, plan and complete tasks on its own initiative.
SaaS moved software into the cloud; Agentic AI gives software a brain.
This transformation is upending three core logics of the software industry:
Pricing logic: from charging per seat to charging per outcome;
Ecosystem logic: companies must control the right to define the semantic layer of their data, or platforms will siphon off their value;
Standards logic: competition such as that between the MCP and A2A protocols is deciding who becomes the "grammar gatekeeper" of the AI world.
Disruption is mandatory. Obsolescence is optional.
Disruption is inevitable; becoming obsolete is a choice.
Compute, energy and machines: AI's "physical bottleneck"
AI's ambitions are putting pressure on the planet's power systems.
Bain estimates that by 2030, US demand for AI compute will reach 100 gigawatts — the equivalent of adding an "AI continent" to the national grid. Building these data centres will require US$500 billion of investment a year.
Even if companies reinvested all their IT savings and AI gains, there would still be an US$800 billion gap.
Meanwhile, humanoid robots and quantum computing are becoming AI's "extended limbs":
Humanoid robots: capital is enthusiastic, but limited by battery density and operational complexity, near-term deployment will remain in factories and warehouses;
Quantum computing: a potential market of US$250 billion, but still more than a decade away from reliably replacing conventional computing.
AI is not just an algorithmic revolution but a restructuring of energy and physical infrastructure.
From pilots to systems: the real watershed of AI transformation
Bain's final conclusion is worth reflecting on for every company and every family:
Leaders have already lifted EBITDA by 10%–25% through AI, while companies still "piloting" are falling dangerously behind.
This means AI competition long ago stopped being about "whether you've deployed it"; it's about whether you've completed the reshaping of organisational structure, data, architecture and talent.
From top-level decision logic to bottom-level data governance, everything needs an "AI-native" redesign.
What the winners have in common:
they use AI to transform processes, not to decorate them;
they treat AI as a factor of production, not an outsourced tool;
they see speed of innovation as a threshold for survival, not a metric for awards.
The operating system of a new civilisation
AI competition has shifted from single-point innovation to system-versus-system rivalry;
it tests not only compute and capital, but even more the speed of cognition and the resilience of organisations.
For family capital, this means:
the core of intergenerational succession is no longer the continuation of wealth, but the ability to understand the new structure.
Whoever first understands the structural forces of AI and positions for them will keep their voice in the new era.
The future won’t wait for you to be ready.
The future will not wait until you are prepared.