Opinion: Huawei's secret weapon wasn't the CCP. It was IBM

• Eric Xu credits IBM with transforming Huawei's approach to innovation
• The same system that helped Huawei win in telecom infrastructure positions it to dominate the world’s emerging AI industrial economy
• Huawei spent $600 million with IBM. Now, America wishes Big Blue had “kept the change”

How did Huawei become one of the world's most pervasive technology companies?

The conventional explanation, particularly in Washington, involves a three-letter acronym: the CCP (Chinese Communist Party).

But in an article for Fierce Network, Huawei's rotating chairman, Eric Xu, credits another three-letter organization with helping drive its success.

No, not the CIA.

IBM.

And two decades later, plenty of American technology companies could benefit from following the same roadmap (including, ironically, IBM itself).

Xu joined the company in 1993 and participated in its transformation from a manufacturer of telephone switches into a global technology powerhouse. The last time I chatted with him, in 2015, he kept his cards close to his chest. Which makes his unusually candid account of Huawei's internal workings, and particularly its relationship with IBM, all the more remarkable.

It's unclear whether this signals a move to a more open communications policy or reflects an effort to shape perceptions amid the current trial between Huawei and the DOJ in a Brooklyn courthouse. Xu may care about neither, given the accelerating pace of Huawei’s ascendancy.

A remarkable rise

Today, Huawei operates in more than 170 countries and regions. Its position is particularly formidable in sub-Saharan Africa — for my American readers, that's the gigantic landmass below the Sahara Desert — where it has become the de facto choice for much of the region's communications infrastructure.

Huawei also has more than 260 patent licensing agreements, embedding its intellectual property in billions of smartphones and millions of connected vehicles, including those sold in countries that ban Huawei's own equipment.

These are extraordinary achievements, especially given the U.S. government's full-spectrum opposition to Chinese technology.

To understand the genesis of this success, we need to rewind to 1997, when Huawei's founder, Ren Zhengfei, visited American technology companies to study their product development methods. He returned convinced that Huawei needed something more valuable than talented engineers. It needed a repeatable system for turning engineering into commercially successful products.

So Huawei hired IBM to build it one, via the Integrated Product Development (IPD) strategy.

IPD was a serious program, with a serious fee to match. As Xu wrote, “From 1997 to 2006, Huawei paid IBM nearly 600 million USD in consulting fees.” For a company generating just half a billion dollars in annual revenue when the relationship began, this was an extraordinary commitment by Ren. Huawei was effectively betting the company on an American consultancy teaching it how to build a better Chinese technology business.

But, boy, did IBM ever deliver on its end of the deal.

IBM helped reorganize Huawei’s research investment, engineering processes, product planning and commercial execution.

But more than that, it reversed the polarity of the Chinese approach to product development, making the systematic identification of customer needs job number one. Rather than squandering R&D budgets on technological vanity projects — engineering marvels conceived to impress other engineers — Huawei dedicated itself to discerning what customers actually wanted, needed and were prepared to pay for, and then making those products.

No more tech for tech’s sake — from that point onward, Huawei would be customer first and last and always.

A study in contrasts

What makes the Huawei-IBM partnership more than a counterintuitive business case is the contrast it presents with the direction of U.S. technology development.

At the same time as Huawei was reinventing itself as a company of customer-driven zealots, the axis of technology development in North America was moving in precisely the opposite direction, as Silicon Valley embraced a venture capital philosophy that regarded customers as people who didn't yet know what they wanted, but jolly well needed to be told. The prevailing conviction was that sufficiently clever VCs, armed with enough capital, could anticipate, shape and accelerate demand before customers themselves knew they needed it.

This turned out to be a spectacularly expensive blunder, producing two overlapping speculative bubbles — the dot-com boom and the optical networking frenzy — whose collapse destroyed billions in investor capital and helped wipe out several North American companies that might otherwise have gone on to put up a decent fight against Huawei.

The asymmetry is stunning. While Silicon Valley's financiers were betting fortunes attempting to will markets into existence through the power of ego, on the other side of the world Huawei was reorganizing itself around customers and markets that already did.

The bigger they are …

I bring all this up not merely because it's interesting (although it certainly is) and because it annoys American investors (same), but because America appears poised to repeat the same cycle of cart-before-horse investment — but this time on a vastly greater scale, with considerably more money at stake and a financial system far more exposed to the consequences.

Because around 2010, a new and more muscular combatant began pushing its way into Silicon Valley's longstanding war against economic logic: private equity. Its practitioners shared the venture capitalists' conviction that enough money could make almost any vision of the future come true. But they brought considerably bigger checkbooks and a rather more dangerous weapon. Where venture capitalists had traditionally gambled with investors' cash, private equity firms had perfected the art of gambling with borrowed money.

Tomorrow's money, in other words. Debt.

Ugly America

The ultimate expression of this philosophy is now rising across America: enormous AI factories, speculative computing carbuncles blighting the landscape, bankrolled by an army of private-equity parasites and complicit institutional money managers, chanting the liturgy of “if we build it, they will come,” while knowing full well the paying customers will never show up in the numbers required to justify a trillion dollars of data center construction.

This is a miscalculation for the ages, right up there with offering subprime mortgages to people without the income to repay them. It substitutes fee collection for financial discipline, passing the risk down the line while counting on the next greater fool to materialize before the debt comes due.

Data centers and AI factories are necessary to an AI economy, sure. But they are only part of its infrastructure, and building a data center does not automatically generate the demand needed to pay for it.

America’s financiers and tech oligarchs defend this data center fool’s gold rush by pointing to the country’s lead over China in AI. But that claim exposes the central problem: they’re talking about frontier models from companies like OpenAI and Google, gigantic information gravity wells powering predictive text engines and chatbots cosplaying as sentience.

Essentially, they wish to treat leadership in one form of AI as leadership in the entire AI economy — like assuming that because you’ve built a machine that writes bad poetry, you can trust it to run air traffic control.

Meanwhile, another AI economy is taking shape, built around machine learning in the physical world. Its geography is different: intelligence close to the machinery, at the edge rather than concentrated in giant data centers. Its engineering requirements are different: task-specific models working within reliable, deterministic control systems, rather than probabilistic language generation. And its purpose is different: making machines work rather than making machines talk. That economy requires coordinated investment in power, communications, computing and industrial operations. A bigger language model does not supply those missing pieces.

If America has demonstrated anything in the last 50 years, it is that it is extraordinarily good at doing “stuff” and extraordinarily bad at fitting said stuff into a sensible strategy.

It can develop extraordinary medical treatments without organizing an effective healthcare system, sustain a war in Afghanistan for 20 years without securing a durable political outcome, and build world-leading internet companies while marooning rural communities in the narrowband boondocks. Now it proposes to build the world’s most expensive computing infrastructure while leaving the electricity, water, communications networks and industrial economy it must serve in separate policy silos.

Huawei, meanwhile, has assembled an integrated infrastructure stack designed to connect computing intelligence to industrial operations. That gives China, and the other 169 countries deploying Huawei solutions, a route from AI investment to machine productivity that America’s frontier-model focus, by itself, cannot match.

And here’s the final irony: one of the most consequential contributions an American technology company made to the next global digital industrial revolution may have been teaching Huawei how to get there first.

America taught Huawei how to run a company, then forgot to run the country.

Learn more about America's and China's technology strategies on Fierce Network

Opinion: America’s big tech exceptionalism risks becoming global ostracism

Opinion: Who will control the control plane of the world?

Opinion: A new era of unified infrastructure

Opinion: My cunning plan to solve America’s AI computing shortage