Whenever one of his direct reports comes in with good news, Intel CEO Lip-Bu Tan interrupts. He explains that what he really wants to hear is the bad news. Then he lays out a three-step scale that doubles as a warning: if you tell me about the problem, it is our problem; if you do not tell me, it is your problem; if I hear about it from the customer, you are in serious trouble.
Tan has been repeating some version of that line for more than fifteen years, across two different companies that he found in trouble. Behind it is a fairly simple observation about how information moves through large organizations: good news travels at the speed of light; bad news crawls. Much of the way he runs a company is designed to correct that imbalance.
What customers said when he asked for a grade
The first time he applied the idea on a large scale was in 2009, shortly after becoming CEO of Cadence, the chip design software company. Instead of announcing a plan, he got on a plane. Korea, Japan, China, Texas. He asked every customer the same question: how would you grade Cadence as a supplier? The grades were bad.
What mattered was not the result but what he did with it. He listened, promised specific fixes and delivered them. Within a year, the relationship had changed. Cadence stopped being treated simply as a vendor and started becoming a partner. One competitor took a while to understand what had happened and eventually admitted it to him: customers treated its suppliers like suppliers, while they opened their five- and ten-year road maps to Cadence. That access to information explains a large part of the recovery. During Tan’s thirteen years as CEO, the stock doubled in value six times.
Inside the company, he followed the same rule. When executives came in talking about what was working, he asked them to start with what was not.
The same scene, sixteen years later
At Intel, he repeated the move on his first day. His largest customer asked him to dinner and arrived with a long list of complaints. Tan pulled out a notebook and started writing.
The response he got was less a compliment than a diagnosis of the previous management: they were pleased that someone was taking notes instead of lecturing them. Today, that relationship is one of the company’s central partnerships.
It is the same playbook he had already used before, including during the interim role that started everything. His move into executive leadership had been accidental. Tan was a Cadence director when the company parted ways with its CEO in 2008, leaving him in charge while a search firm looked for a replacement. He threw himself into the job anyway. Two weeks later, the chairman asked the search firm to add Tan to the candidate list. Two weeks after that, he told them to stop looking.
An engineering problem, even when the problem is the president
Tan’s background explains a great deal about how he approaches problems. He was born and raised in Malaysia, where he studied physics. At 19, he moved to the United States and earned a master’s degree in nuclear engineering from MIT. He later added an MBA from the University of San Francisco and, at 27, founded Walden International, a venture capital firm that backed semiconductors when much of the industry’s investment money was looking elsewhere. Over the years, it supported hundreds of U.S. companies that generated roughly 50,000 jobs and nearly US$400 billion in combined market capitalization.
That mindset showed up during one of the most uncomfortable moments of his first year at Intel. One Thursday at 5:30 in the morning, as he was getting ready for his gym and swimming routine, his phone would not stop ringing: Donald Trump was publicly demanding his resignation over an alleged conflict of interest, leaving little room for nuance.
Tan tried to remove himself from the equation. He did not need the job and, as he tells it, had taken it to fix the company. From there, he treated the situation like any other failure that needed to be solved. The first item on the agenda was simple: get a meeting.
He called four friends with access to the White House and was received on Monday. He told his story, pointed out that Cadence’s recovery had taken place during Trump’s first term and offered a slogan tailored to the second: make Intel great again. When asked how he planned to achieve what others had failed to do, he laid out the plan. He was invited back the following Friday. The government wanted 10% of the company.
Tan defends government ownership without making it sound extraordinary. TSMC began with the Taiwanese government as a shareholder, and similar models exist in Japan and Singapore. To him, it is infrastructure, not an exception.
Fix the balance sheet first, because nobody wanted to join
Intel’s first priority under Tan looks financial, but it was also about people. He spent the first six months repairing a damaged balance sheet after discovering that he could not recruit without doing so. The best candidates were turning him down because, from the outside, Intel looked almost bankrupt. After raising US$20 billion, those conversations started to change.
Once that was under control, a long and unglamorous list of jobs followed. Simplify a product line that had become unmanageable. Redesign the chip design methodology because too many chiplets were slowing down the response to customers. Regain lost ground in data centers and restore features that had been abandoned without a convincing reason, such as multithreading. And move beyond selling silicon toward offering a full platform, because customers no longer ask for a single chip: they want the entire rack and the software layer behind it.
In leading-edge manufacturing, his argument starts with one number: 97% of the world’s high-performance silicon comes from a single company outside the United States. Intel, Tan argues, is the only company capable of both designing and manufacturing entirely inside the country. But he also warns that technology alone is not enough.
Foundry is above all a business built on trust. A customer is betting billions of dollars in revenue on the assumption that the wafers will come out right. Building that trust, he says, does not happen in one leap. It takes roughly a hundred steps.
How to talk to an agent
The same logic holds when he is asked about artificial intelligence. Tan believes its impact will be ten times greater than the internet’s and does not think expectations are excessive in the short term. But the thought that best reveals how he sees it comes from home.
He has three grandchildren between the ages of three and five, and he has been talking with his son and daughter-in-law about how to prepare them for the world they will face twenty or thirty years from now. Their conclusion was not to choose a technical profession. Tan believes what will matter is learning how to communicate with agents and digital workers, how to coordinate them and how to exercise judgment. Once again, the same principle appears: listen well and know what to ask.
At 66, after several friends advised him not to take the job —“people only remember the last one,” they warned him— Tan made it clear from the beginning that this would not be a one-year turnaround. He expects it to take at least five. He describes the current year as the year of execution and does not expect growth until the next one, after five years of flat or declining numbers.
Before accepting, there was still one negotiation left. It was at home. Permission finally came on his 44th wedding anniversary, when his wife told him to do it.
