NextFin News - Forrester Research founder and CEO George Colony is sizing up the AI boom at the moment the industry is split in two. Nvidia is placing a $3.5 billion bet on chip partner MediaTek, investors are gearing up for a potentially record-breaking Anthropic initial public offering, and yet the research firm Colony runs is forecasting that enterprises will defer a quarter of their planned AI spending into 2027. The tension between the infrastructure frenzy and the enterprise pause is the real story.
The question the market has not fully priced is not whether artificial intelligence is transformative - Colony has called generative AI "the biggest technology change of my lifetime" - but where in the stack the next leg of returns will come from, and how long the current reckoning over return on investment lasts. His framework, drawn from Forrester's own forecasts and his public writing, points to a boom that is structural in the plumbing and cyclical in the boardroom.
The Setup: A Boom With a Reckoning Attached
On Aug. 31, 2026, Colony weighed in during a television interview on the AI boom, whether the hype is justified, and what comes next. The backdrop was itself a study in extremes. Nvidia announced a $3.5 billion investment in MediaTek, deepening its supply-chain entanglement as demand for AI silicon shows no sign of cooling. Anthropic, which raised $65 billion at a $965 billion post-money valuation in late May 2026 and disclosed roughly $47 billion in run-rate revenue by mid-year, has confidentially filed for a U.S. listing that could rival the largest technology debuts on record.
Against that backdrop, Forrester's own research delivers the counter-melody. In its 2026 technology and security predictions, the firm argued that "the AI hype period ends" in 2026 as pressure to deliver measurable results intensifies. The mechanism is blunt: fewer than one-third of decision-makers can tie the value of AI to their organization's financial growth, so CEOs will lean harder on CFOs, financial rigor will slow production deployments, and enterprises will defer 25% of planned AI spend into 2027. Sharyn Leaver, Forrester's chief research officer, put it directly: "In 2026, the AI hype period ends as the pressure to deliver real, measurable results from secure AI initiatives intensifies."
And the messenger is living its own message. Forrester's second-quarter 2026 revenue was $100.2 million, down from $111.7 million a year earlier, with contract value at $283.2 million, down 3%. For full-year 2026, management guides revenue of $345 million to $360 million - down 9% to 13% versus 2025 - as it sunsets its strategy-consulting line and reshapes the business around research, which is expected to rise to roughly 80% of total revenue. Colony has said Forrester is "uniquely positioned to help large companies navigate these problems," attacking the AI opportunity from two directions: using AI to improve how it delivers services, and selling AI-focused research.
So the central tension is not abstract. The firm whose CEO is being asked to size up the AI boom is itself in the middle of an AI-forced transition - revenue falling, product lines closing, staff reduced. That makes Colony's framework worth parsing carefully, because he is not selling the boom; he is describing a wave that drowns some businesses while lifting others.
Colony's Framework: The Seventh Wave, Not Another Cycle
Colony has worked through six major technology shifts - minicomputers, the PC, the internet, social, mobile, and cloud - and he frames AI as a "Seventh Wave" of comparable magnitude. That framing is not neutral. Calling something a "wave" is a structural claim: the change will not mean-revert, a new set of providers will rise and swaths of legacy companies will be destroyed, and the old defensive playbook - buy the interlopers, block with regulation and packaging, pretend to be part of the new era, or link new offerings to dominant existing products - will only slow, not stop, the transition.
"This is an 'Iron Man' moment, not a robot moment - it's an opportunity to put workers in 'suits' of technology to enable them to serve customers better," Colony wrote.
The metaphor matters. A robot replaces the worker; a suit amplifies the human wearing it. Colony's framing implies that AI's economic payoff arrives through augmented labor, not headcount elimination alone - and that the winners will be the firms that redesign workflows around the suit rather than bolt the technology onto existing processes. He has even coined a term for it: "Human AI," combining digital speed with human connection.
His sector-level calls give the framework concrete shape. Hardware, he expects, will grow in the 8% to 10% range per year over the next five years, though "the NVIDIA stranglehold will take another 12 to 15 months to break." Technology services, battered since 2023 by the overexpansion of 2021-2022, could grow 5% to 6% annually from 2026 to 2030, up from Forrester's 3.6% estimate for 2025, as firms such as Cognizant and Capgemini help clients migrate away from legacy systems. Telecommunications and communications-equipment growth - 1.5% and 0.8% globally in 2025 - "could be doubled" by the movement of prompts and answers between users and AI data centers, because at least 70% of AI computing will run off private and public clouds rather than on edge devices.
That last number is the hinge of the whole argument. If 70% of AI computing stays in the cloud, then the spending wave is not just about chips. It is about data centers, networking, power, cooling, and the services layer that ties them together. The hardware boom has a longer runway than a single product cycle would suggest.
The Numbers: Growth in the Aggregate, Pain in the Mix
Forrester's market forecast for 2026 captures the duality. Global technology spending is projected to grow 7.8% to $5.6 trillion, up from $5.2 trillion in 2025 - a record pace despite U.S. tariffs and economic volatility. Michael O'Grady, a principal forecast analyst at Forrester, attributed the momentum to "continued investment in and adoption of AI" across defense, financial services, healthcare, industry, and retail. Nearly two-thirds of global tech-spending growth over the next five years is expected to come from software and computer equipment, especially servers. Hyperscalers are forecast to capture almost half of AI infrastructure capital expenditure between 2025 and 2028, while infrastructure-as-a-service is expected to compound at 22% annually from 2024 to 2030. By 2030, AI-impacted hardware is projected to capture more than 80% of computer-equipment spend, up from 43% in 2024.
Read that sequence carefully. The aggregate number - 7.8% global growth - is the headline that bulls will cite. But the composition is doing the real work: spending is concentrating in infrastructure and cloud, while the applications and consulting layers face the ROI reckoning that Forrester's predictions describe. That is why the same firm can simultaneously forecast record tech-spending growth and a 25% deferral of planned AI spend. The money is not leaving the ecosystem; it is moving up the stack and down the timeline.
There is a second, quieter number worth holding next to the first. Forrester's own revenue is guided down 9% to 13% in 2026 even as it forecasts 7.8% growth for the global tech market. The divergence is not a contradiction; it is the point. A research-and-advisory firm selling strategy consulting is exposed to the exact budget line - discretionary, hard-to-measure AI programs - that CFOs are now scrutinizing. Colony's pivot toward research and AI-enabled delivery is a bet that the durable part of the AI trade is the part enterprises cannot postpone: the infrastructure buildout and the governance required to run it.
The productivity evidence already on the ground shows why the reckoning is selective rather than total. Forrester's spending forecast cites Microsoft's disclosure that as much as 30% of its code is now written by generative AI, and JPMorgan Chase's experience of flat fraud-detection costs despite a 12% annual increase in threat attack rates, alongside a 10% to 20% rise in software-engineer productivity after investing in AI. The technology is delivering. The question is who captures the value - and whether the enterprises writing the checks can prove it landed in their own P&L.
Second-Order Read: The Boom Is Structural, the Spending Wave Is Cyclical
The market's dominant narrative treats the AI boom as one thing - either a durable regime shift or a bubble about to pop. Colony's framework suggests both are true at different layers, and confusing them is where investors make mistakes.
The structural leg runs through the plumbing. Cloud compute, networking, power, and the migration of legacy systems are not discretionary in the way a chatbot proof-of-concept is. Once an enterprise commits to running 70% of its AI workloads off cloud infrastructure, that spend behaves like a utility buildout: it front-loads capital expenditure and then persists. Colony's 12-to-15-month timeline for the Nvidia stranglehold to break is itself evidence of structural thinking - he is not predicting Nvidia's decline, but the predictable diversification that follows any platform monopoly. The hardware growth call of 8% to 10% annually for five years is a structural claim, not a cycle call.
The cyclical leg runs through the boardroom. The 25% deferral of AI spend, the CFO's new veto power, and the end of the "hype period" are classic late-cycle behaviors: after a capital rush, the entities writing the checks demand proof of payoff. This is mean-reverting by construction. Deployments paused for ROI review do not get cancelled forever; they get repriced, re-scoped, and pushed into 2027. The cycle trough is the reckoning year. The cycle peak was the unfettered budget approval of 2024-2025.
The second-order implication is where the conventional wisdom breaks down. The consensus read of "AI spending slows" is that the boom is rolling over. The second-order read is that a spending slowdown at the application layer is bullish for the infrastructure layer's pricing power and bullish for the services and governance layer's margins - because the money that does get spent is concentrated in projects with executive sponsorship and measurable returns. Scarcity of approved capital raises the hurdle rate, and the hurdle rate is a filter, not a shutdown. The firms that survive the reckoning are the ones Colony's Iron Man metaphor describes: those that put the suit on and redesign the work, not those that bought licenses and waited.
There is also a cross-asset transmission most investors are underweighting. If enterprises defer 25% of AI spend into 2027 while hyperscalers keep committing to half of AI infrastructure capex through 2028, the duration of the AI trade lengthens even as its near-term growth rate moderates. That favors capital-intensive, long-duration assets - data-center real estate, power infrastructure, networking equipment - over short-cycle software experiments. The boom does not end; it migrates into assets with longer payback periods.
The Counter-Thesis: What If the Reckoning Is the Beginning, Not a Pause?
The strongest case against Colony's framework is not that AI fails to deliver value - the productivity evidence is already visible in code generation, customer support, and fraud detection. It is that the ROI gap is not a measurement problem that better governance will solve, but a fundamental economics problem: the value created by AI accrues to consumers and to a narrow layer of infrastructure providers, while the enterprises buying the technology absorb the cost. In that world, the 25% deferral is not a cyclical pause but the first sign of a permanent compression in enterprise AI budgets, and the "hype period ending" is a polite way of describing demand destruction.
This view has credible backing. It is the position implicit in every CFO who now requires a financial tie-out before approving an AI project, and it is consistent with the pattern of previous general-purpose technologies, where the bulk of the economic surplus flowed downstream to users rather than upstream to the technology vendors. If it is right, then Colony's structural claims about hardware and cloud are correct but insufficient: the infrastructure buildout would overshoot the revenue that the application layer can ultimately support, and the 8% to 10% hardware growth call would prove to be a mid-cycle peak rather than a five-year run rate.
The answer to that counter-thesis rests on the composition of spend, not its level. Infrastructure demand is not solely a function of enterprise application ROI; it is also driven by sovereign AI programs, defense spending, and the competitive arms race among the hyperscalers themselves - none of which require a CFO's sign-off on a specific use case. That is why the deferral forecast and the infrastructure forecast can coexist. But the counter-thesis identifies the real vulnerability: if hyperscaler capital expenditure is ultimately funded by application-layer revenue that never materializes, the capex cycle will turn, and the duration trade will reverse.
The falsifying signal is specific. If, by the end of 2027, enterprise AI spend has not recovered from the 25% deferral and hyperscaler AI infrastructure capex growth falls below 15% year over year for two consecutive quarters, then the structural-plumbing thesis is wrong and the boom was predominantly a cyclical capital-expenditure impulse. Watch the hyperscaler capex disclosures and Forrester's own 2027 predictions update - those are the tripwires.
What Comes Next: Three Horizons, Three Scenarios
Short term (the rest of 2026): sentiment is dominated by the reckoning. Expect continued pressure on discretionary AI budgets, more CFO-led reviews, and a bifurcation between infrastructure suppliers - who keep pricing power - and application vendors selling unproven productivity gains. Forrester's own revenue guidance, down 9% to 13%, is a leading indicator of this pressure in the advisory and consulting channel.
Medium term (2027-2028): the deferred spend returns, but in a different shape. Projects that survive ROI review are larger, better governed, and tied to measurable outcomes. The services layer - migration, integration, governance - captures a larger share of the wallet than it did during the hype phase. This is the window where Colony's 5% to 6% tech-services growth call would be validated or broken.
Long term (2029-2030): the structural claims are tested. If 70% of AI computing remains cloud-based and AI-impacted hardware reaches the projected 80% of computer-equipment spend by 2030, the Seventh Wave framing holds and the current reckoning will be remembered as a mid-wave consolidation. If instead edge deployment accelerates faster than expected and enterprise budgets compress permanently, the wave was shorter and shallower than Colony's six prior analogs suggest.
Base case: infrastructure keeps building through 2028, enterprise spend returns in 2027 in a more concentrated form, and the advisory firms that pivoted to research and AI governance - including Forrester itself - stabilize as the market moves from buying tools to buying outcomes. Upside case: the Anthropic listing and continued hyperscaler commitment extend the capex supercycle beyond 2028, lifting the entire stack. Downside case: the ROI gap proves structural, hyperscaler capex rolls over in 2027, and the hardware growth call compresses to mid-single digits.
The AI boom is real, but it is not one boom. The structural boom - in cloud, power, networking, and the rewiring of enterprise workflows - is only in its second inning. The cyclical boom - in unfettered application spending - has already peaked, and its hangover is the reckoning Colony is describing. Investors who treat the two as the same thing will sell the infrastructure trade at the bottom of the application cycle and buy the application trade at the top of the hype cycle. The suit is real; not everyone wearing it is Iron Man.
Explore more exclusive insights at nextfin.ai.
