Bridging the AI Velocity Gap

By Ravi Kumar S, Andreea Roberts and Simone Crymes

Cognizant

ShareNewsweek is a Trust Project member

The AI market is caught between two clocks running at very different speeds. One clock, representing the infrastructure build-out, is spinning quickly as innovation shifts the frontier of hardware and AI models every six to 12 months, compressing the payback window.

The other clock, representing the diffusion of AI inside enterprises, is spinning slowly. Businesses struggle with an elongated J-curve in which initial investments and adoption realize modest returns and the path to significant growth—via new products and services unlocked by business reinvention—seems distant.

...

This speed gap creates a significant challenge to both the companies building out the infrastructure and the companies looking to adopt AI, mirroring several historical episodes of rapid investment, overcapacity and slower-than-expected adoption that caused short-term, but acute, economic pain.

The fastest way to close this gap is to help businesses compress the J-curve by accelerating AI absorption in the enterprise. A new kind of company—the “AI builder”—is emerging to do just that.

The Race Against Time

We’re in the middle of the largest infrastructure bet in tech history. Hyperscalers and large technology platform companies are investing hundreds of billions of dollars a year into AI infrastructure. By 2030, global spending on AI data centers, chips and power is expected to exceed $5 trillion. To justify this investment, AI will have to generate multiples of this figure in value for end-user industries.

Concerns of a potential overbuild are growing stronger among market participants because many believe they’ve seen this story before. Most recently, the late-1960s was marked by massive capex on fiber optic cable and network equipment in anticipation of internet usage growth. The growth came, but slower than expected, and as a result, created a wave of destruction (Global Crossing, Worldcom, et cetera) as the market waited for utilization to catch up with expectations.

Walk back in time and history is littered with similar episodes: satellite communications in the 1960s and 1970s, mainframe computing in the 1950s and 1960s, radio and broadcast infrastructure of the 1920s and 1930s, and electrification at the start of the 20th century. In each, general-purpose technology innovation followed a pattern: infrastructure materialized rapidly as companies piled in to capture expected returns. But, in doing so, they overestimated the speed with which end-users would adopt the new technology. The result was overcapacity and economic pain for investors in the short-term and a meaningful delay in economic benefits for adopters.

All of these instances eventually resulted in widespread adoption of the new technology and enjoyment of the corresponding economic benefits. For example, the fiberoptic cable initially seen as waste eventually provided cheap capacity that enabled cloud computing and the globalization of work and the services value chain.

However, there are two key differences between the current AI infrastructure investment boom and these historical episodes. The first is the rapidity of innovation. The frontier of hardware and LLMs is evolving every six to 12 months, as opposed to the much slower pace of technological progress experienced in the past. The result: investment obsolescence is a much greater risk today.

Second, past technology investments were often financed by borrowing, and in doing so, explicitly imposed a timetable on investment returns with the maturity of the debt. AI infrastructure investment is being financed with operating cashflows with no explicit deadline. Nonetheless, the owners of these cash flows, debt and equity investors still enforce a timetable with their ability to reallocate capital if expected returns don’t materialize quickly enough.

In sum, time, not capital, is the binding constraint today. AI must yield value quickly.

AI Builders as the Bridge

JP Morgan estimates that, for a 10 percent return on current infrastructure investments, the AI infrastructure ecosystem needs roughly $650 billion in new annual revenue. That bill ultimately falls to customers—a mix of enterprise spending and consumer-facing products. But the sticker price is just the beginning; enterprises must achieve multiples in cost savings and revenue which, once integration, agent development and operations are added, implies trillions of dollars of value creation needed in end-user industries like banking, insurance, retail, manufacturing and health care.

Enterprises are eager to realize the full promise of AI, yet most are moving through a J-curve. Legacy workstreams are being automated to re-baseline productivity; however, the next phase of growth—new revenue streams and new jobs rooted in true business reinvention—is not yet visible.

At the start of the digital era, complete business transformation took anywhere between three and seven years at the firm level, and increased productivity did not show up in GDP figures until more than a decade after the arrival of the PC.

While AI’s adoption curve is likely to be much steeper than previous general-purpose technologies, the complexity and risk associated with implementing it at scale are unprecedented, creating real friction and slowing down the take-off phase.

“AI builder” companies are emerging to bridge the velocity gap between AI infrastructure investments and business value realization—compressing the J-curve. These companies are building the last mile of enterprise adoption to operationalize AI at scale. They’re helping convert generic AI infrastructure into business outcomes through platforms and services that are contextualized to each business and help shield enterprises from the underlying complexity and infrastructure churn.

What matters now is conversion speed: turning AI capacity into operating results. AI builders accelerate that conversion by derisking and industrializing the last mile, the only way to translate the $5 trillion AI infrastructure investment into durable value.

Ravi Kumar S is the chief executive officer of Cognizant leading 350,000 associates and $20 billion in annual revenues, partnering with Global 2000 Companies in their AI journeys and reimagining enterprise and workforce landscapes.

Andreea Roberts leads technology marketing globally at Cognizant, helping clients convert new technologies into market-shaping advantage.

Simone Crymes is chief of staff to Cognizant’s CEO, building bridges across strategy, technology and human impact to create meaningful enterprise transformation.

The views expressed in this article are the writers’ own.

Request Reprint & LicensingSubmit CorrectionView Editorial & AI Guidelines