Goliath’s dilemma: Can Big Tech be disrupted, or will the giants always win?

Every few years someone declares that the age of Big Tech dominance is ending. A new startup has cracked the code, a regulator has found its courage, a geopolitical rival has closed the gap. And every few years the giants absorb the shock, buy the challenger, or lobby the regulator into a holding pattern. The cycle has repeated often enough that scepticism about disruption is now the default position among serious observers of the technology industry.

But the landscape in 2025 and into 2026 looks different from previous cycles in ways that are worth examining carefully. Artificial intelligence has created enormous new concentration of power among the largest firms, while also opening competitive surface area that genuinely did not exist three years ago. The lobbying budgets are bigger than ever. The acquisitions are more expensive and more creative in structure. The regulatory response is more coordinated across jurisdictions than at any previous point. And the talent is moving, out of the incumbents and into startups, at a pace that tells you something about where the people closest to the technology think the real opportunities now sit.

The question of whether Big Tech can actually be disrupted matters well beyond the technology sector. It touches industrial policy, competition law, national security, and how democratic societies govern concentrations of private power that now rival some state institutions in reach and influence. The combined market capitalisation of the five largest US technology companies exceeds the GDP of every country on earth except the United States and China. These are not ordinary corporations competing in ordinary markets. They operate critical infrastructure, control the platforms through which public debate takes place, and increasingly shape the policy environments meant to regulate them. The data from the past eighteen months gives us a sharper picture than we have had in a decade of where the real advantages sit, where they are weakening, and what the most probable competitive landscape looks like going forward.

What the incumbents are actually working with

Start with capital, because everything else flows from it. Apple, Microsoft, Alphabet, Amazon, Meta, and Nvidia generate hundreds of billions in annual free cash flow. SoftBank committed $40 billion to OpenAI in 2025. The Stargate Project reached $500 billion in announced investment. Nvidia sold more than $50 billion in data centre chips in a single quarter. Training a frontier large language model now costs hundreds of millions of dollars, approaching a billion for the largest runs. These numbers have a weight to them that shapes everything downstream.

When the internet took off in the 1990s, two people in a university room could build a search engine. Cloud computing in the 2010s made infrastruture even cheaper to access. AI has reversed that direction completely. The compute, energy, and data requirements of frontier model training are so large that only a handful of organisations on the planet can credibly operate at the cutting edge. Some analysts now question whether even OpenAI, the company most associated with the current AI wave, can sustain its spending without borrowing against uncertain future revenues, while Google, Meta, and Microsoft fund equivalent efforts from cash flow they already have. The United States recorded about 28 gigawatts of data centre load in 2025, expected to exceed 60 by 2030. In Virginia alone, the Colossus complexes secured over one gigawatt of dedicated gas generation. Securing land, power, and cooling at this scale takes years of planning. Incumbents who started investing a decade ago hold positions that no new entrant can replicate quickly.

Then there is the political machinery. Seven of the largest technology and AI companies spent a combined $50 million on federal lobbying during the first nine months of 2025. That works out at nearly $400,000 for every day Congress was in session. Meta spent a record $13.8 million in the first half of the year and employed 86 lobbyists, about one for every six members of Congress. ByteDance, fighting for TikTok’s survival, maintained 42 lobbyists. In Brussels, the digital industry spends approximately €151 million annually on EU lobbying, up more than 50 per cent from four years earlier, with Big Tech averaging more than one meeting per working day with senior Commission officials. The top ten digital companies outspend the top ten in pharma, financial services, and automotive by a factor of three.

Public Citizen’s analysis found that Big Tech executives and investors committed at least $764.5 million during the 2024 election cycle and through 2025, with nearly three quarters favouring Republicans. Meta launched multiple super PACs. OpenAI co-founder Greg Brockman put $100 million into a new political action committee. The strategy has shifted from shaping how laws are written to determining who writes them. In 2025 the industry pushed a provision in the US spending bill that would have blocked states from regulating AI and social media algorithms for ten years. It was stripped out after civil society pushed back, but the fact it got that far tells you something about the reach. In Europe, US Vice President JD Vance publicly attacked European free speech protections and Secretary of State Marco Rubio urged diplomats to undermine the EU’s Digital Services Act. The line between corporate lobbying and state foreign policy has blurred to the point where its hard to say where one ends and the other begins.

And then there is the acquisition playbook, which remains the most direct way to neutralise a competitive threat. Facebook bought Instagram for $1 billion in 2012 when it had 30 million users and 13 employees. WhatsApp went for $19 billion in 2014. Neither deal was seriously challenged at the time. Google’s $32 billion acquisition of Wiz in early 2025, its largest ever, came after Wiz rejected a $23 billion bid in mid-2024 partly over concerns about Biden-era regulatory scrutiny. Eight months and a change of administration later, Wiz said yes to a higher number. Meta took 49 per cent of Scale AI for $14.3 billion. Palo Alto Networks bought CyberArk for $24.5 billion. Tech M&A rose 36 per cent by value in 2025, with more than five deals above $10 billion. Global M&A hit $4.8 trillion for the year, the second-highest on record, with technology leading all sectors.

The AI era has added a new move to the playbook. When OpenAI’s attempted purchase of coding startup Windsurf fell apart because of conflicts with Microsoft, Google simply hired Windsurf’s CEO and key engineers while licensing the technology, in a package valued at around $2.4 billion. No formal acquisition. No merger review. The acqui-hire structure has become standard for absorbing startup innovation at speed, and regulators are only beginning to grapple with whether it amounts to the same thing as a takeover by another name. Many of the headline AI transactions in 2025 were built around acquiring a small number of elite researchers and founders at premium valuations. The European Commission’s public consultation on updated merger guidelines now explicitly addresses these structures, acknowledging that the acquisition playbook has evolved faster than the rules designed to contain it. The pattern is well worn at this point: a startup gains traction, attracts a giant’s attention, and gets absorbed before it can build the independent scale that would make it a lasting competitive force. For every Instagram that could have become a genuine rival, there are dozens of smaller acquisitions that never attracted public scrutiny at all.

Where the cracks are showing

All of which paints a picture of near-invulnerability. But spend time with the details and a more complicated picture emerges. Large technology companies carry structural disadvantages that capital cannot fully offset. They are complex organisations with multiple product lines, entrenched internal cultures, and powerful constituencies that resist change even when the threat is visible. Microsoft could see mobile computing coming but couldn’t bring itself to cannibalise Windows. Google understood social networking but never seriously challenged Facebook. Meta saw short-form video rising and still responded slowly to TikTok. These weren’t failures of intelligence. They were failures of organisational structure, and they recur because the incentives inside large companies consistently favour protecting what exists over building what might replace it.

Startups don’t carry that baggage. They build for one use case without having to worry about cannibalising an existing product line or navigating seven layers of internal approval. In the current AI landscape this is producing companies that are genuinely difficult for the incumbents to replicate. Cognition AI built Devin, an autonomous AI software engineer that writes, tests, and deploys code, going directly at the developer tools from Microsoft, Amazon, and Google. Together AI is giving developers control over open-source LLM infrastructure that the hyperscalers have kept proprietary. ElevenLabs dominates synthetic voice despite competing against companies with vastly more resources. AI-native startups, those building from the ground up on AI rather than bolting it onto legacy products, can scale fast with very lean teams because the economics of building on existing foundation models have changed what’s possible at the application layer.

The talent picture reinforces this. Several of the most prominent AI researchers have left Big Tech in the past eighteen months to start their own ventures. Ilya Sutskever departed OpenAI in 2024 to found Safe Superintelligence, raising over a billion dollars. Mira Murati left to start Thinking Machines Lab. Yann LeCun, Meta’s chief AI scientist, departed after clashing with leadership over research direction. Meta then laid off 600 people from its Superintelligence Lab in October 2025, including senior FAIR researchers. Jeff Bezos named himself co-CEO of Project Prometheus, an AI manufacturing startup. When the people who actually built the technology inside the giants decide they can do better work outside them, that tells you something the financial statements don’t capture. AI investment exceeded $109 billion in 2024 and a significant share went to startups led by exactly these kinds of people. The World Economic Forum noted in early 2025 that AI-native startups are redefining entrepreneurship itself, achieving scaling velocities with teams that would have been considered impossibly small five years ago. The talent pool gravitates towards major hubs and big-tech salaries, but the founders leaving those very organisations are pulling some of the best people with them.

The geopolitical dimension adds pressure that the acquisition playbook simply cannot address. Huawei’s continued development of competitive telecom infrastructure and smartphone chipsets despite the most severe US export controls ever imposed on a technology company showed the limits of containment. Huawei is no longer just a telecoms company. It underpins energy networks, transport systems, manufacturing platforms, and healthcare across much of the developing world. In ASEAN and the Global South, governments want integrated packages that combine power, connectivity, and digital platforms under a single financing plan and delivery schedule. Huawei delivers that. No Western competitor currently does.

TikTok demonstrated that a Chinese company could build a consumer product compelling enough to force Meta and Google to redesign their own platforms in response. The US government reached for national security legislation rather than antitrust or competitive strategy, which amounted to an admission that the market alone could not contain the threat. TikTok’s US operations were restructured in 2025 under American investors including Oracle, Silver Lake, and UAE-backed MGX, creating a precedent for state-directed intervention in consumer technology that will shape how governments approach foreign-owned platforms for years to come. China’s AI companies, including DeepSeek, have produced competitive large language models at a fraction of American costs. Inside China competition is fierce and margins are thin, which drives companies to export aggressively and bundle infrastructure offerings for international markets. The competitive pressure these firms exert sits entirely outside the reach of Washington lobbyists or Palo Alto law firms. You cannot acquire a company backed by the Chinese state, and you cannot lobby Beijing into regulatory concessions. The Western playbook for managing competition simply does not apply to this category of challenger, and that is a structural vulnerability the incumbents have not yet found a convincing answer to.

The regulatory environment is also more hostile to incumbents than at any point in the modern technology era, despite the lobbying. Courts found Google holds monopoly power in search distribution and ad technology. The DOJ and FTC have active cases against Google, Apple, Amazon, and Meta. The FTC has advanced a broad probe into Microsoft’s cloud, AI, and software businesses. The Amazon case, alleging self-preferencing and algorithmic price manipulation through a system internally codenamed Nessie, survived its motion to dismiss. The DOJ’s case against Apple, alleging it unlawfully restricts cross-platform technologies across its device ecosystem, is also proceeding. In the UK, the CMA designated Google and Apple with Strategic Market Status under the DMCC Act, with a third SMS investigation expected in early 2026. The European Commission is consulting on updated merger guidelines that explicitly address killer acquisitions and digital ecosystem effects, responding directly to the Draghi report’s call for more agile competition enforcement in fast-moving technology markets. None of this is moving as fast as the deals themselves. But the cumulative weight of litigation across multiple jurisdictions constrains what incumbents can do, even where individual cases take years to resolve.

Where this is heading

The honest read is that Big Tech’s advantages are real and durable but not permanent, and the nature of disruption is changing rather than disappearing. The capital intensity of frontier AI strongly favours the incumbents. Political spending is at record levels. The acquisition machinery is well practised and getting more creative. These are not temporary conditions.

But the constraints are accumulating. Regulatory scrutiny across the US, UK, and EU is higher than it has ever been. The talent flow is moving in the wrong direction for the giants. Geopolitical competitors cannot be acquired or lobbied into submission. The capital barrier matters enormously at the frontier of model training but much less at the application layer, in vertical AI, in developer tooling, and at the network edge where local data and domain knowledge matter more than raw compute. Sovereign wealth funds are becoming more assertive in technology investment, potentially funding alternatives that the established venture capital ecosystem might not.

The historical pattern is instructive. Google did not build a better portal than Yahoo. Facebook did not improve on MySpace. The iPhone did not out-feature Nokia. In each case, the disruptor redefined what the category was for rather than competing on the incumbent’s terms. The startups that will matter in this cycle are not trying to out-train GPT or out-scale AWS. They are finding applications and markets that the giants cannot reach because of how their own organisations are wired. The incumbents know this, which is exactly why the acquisition machinery runs at such volume and speed. The race is between the challengers’ ability to build independent scale and the giants’ ability to buy them before they get there.

The UK’s position deserves specific attention. The CMA’s Strategic Market Status powers under the DMCC Act are among the most advanced competition tools available anywhere for addressing digital market concentration. How the UK chooses to use them, assertively or cautiously, will signal to both incumbents and startups how seriously mid-sized economies intend to enforce competition in markets dominated by American and Chinese firms. The window to establish the UK as a credible jurisdiction for technology competition is real, but it won’t stay open indefinitely.

For new entrants the strategic calculus is relatively clear. Competing on infrastructure or frontier model training is a resource contest that the incumbents will win almost every time. Competing at the application layer, in specific industry verticals, and at the edge of networks where execution speed and domain expertise outweigh raw compute is where the structural advantages begin to shift. The companies that define the next decade of technology will most likely be those that understand where the giants are slowest and most constrained, and concentrate their resources there.

The balance of probability

Big Tech’s dominance is not going to collapse. The combination of capital, political reach, and acquisitive capability makes these companies extraordinarily difficult to displace from their core markets. No startup is going to take search from Google or cloud from Amazon through head-on competition.

The most consequential disruptions have never worked that way though. They come from adjacent territory, from categories that didn’t previously exist, from shifts in where value accumulates in the technology stack. The talent exodus, the regulatory tightening, and the geopolitical challengers operating outside the Western acquisition framework all point toward sustained pressure at the margins of incumbent power. The most probable outcome is a gradual rebalancing rather than a dramatic overthrow. Incumbents keep their core platforms. Challengers capture the next layer of value creation. The physical constraints of AI, the energy bottlenecks and infrastructure limits already reshaping where compute gets built and allocated, will reward those who execute fast in specific domains over those trying to control everything from the centre. The edge of the network, where real-world data meets operational systems, is where much of this next wave of value will be created, and it is precisely the territory that centralised platform companies are least well positioned to serve.

Whether this rebalancing actually happens depends less on technology than on political will. The competition tools exist, in Washington, Brussels, and London. The question is whether governments apply them with enough consistency to keep markets genuinely contestable, or whether the lobbying budgets and the political donations do their work and the concentration deepens further. That is a political judgement as much as an economic one, and right now the outcome is far from settled.

illuminem Voices is a democratic space presenting the thoughts and opinions of leading Sustainability & Energy writers, their opinions do not necessarily represent those of illuminem.

Interested in the companies shaping our sustainable future? See on illuminem’s Data Hub™ the transparent sustainability performance, emissions, and climate targets of thousands of businesses worldwide.