Broken CPUs, workforce cuts, cancelled dividends and a decade of borked silicon—how has it all gone so wrong for Intel?

Let's begin with a brief recap of the current state of play at Intel. For starters, despite bold promises to regain technology leadership, it remains miles behind TSMC (Taiwan Semiconductor Manufacturing Company Limited) in chip manufacturing and indeed is increasingly relying on TSMC to manufacture its latest and future CPUs, such as Meteor lake, Lunar Lake and Arrow Lake.

Meanwhile, it's losing market share to AMD in server CPUs, Qualcomm's Snapdragon X Arm-based chips are a real threat in Intel's largest consumer market, laptops, Intel's Arc graphics effort has been a bit of a flop so far, and now its last two generations of desktop CPUs are badly broken.

Most recently, Intel announced some very poor financial results and decided it needed to fire another 15,000 employees after already trimming 5% of its workforce last year, a move that CEO Pat Gelsinger branded «some of the most consequential changes in our company's history.»

That's just the Cliff notes. There are several other lesser horror stories, like reigning in plans to build fabs in Europe and the fact that things are so bad Intel cancelled dividend payments to shareholders. But an obvious question follows. How did Intel get it all so wrong?

Arguably, you could say the rot set in when Paul Otellini replaced Craig Barrett as CEO way back in 2005. Barrett had a PhD in materials science from Stanford. Otellini was an economics grad with an MBA. Put simply, sales replaced engineering at the top of Intel.

Admittedly, Intel's next CEO Brian Krzanich was from the engineering side, but let's not allow details like that to get in the way of an easy narrative. The story goes that Intel got lazy and cynical under sales and marketing-driven leadership, leaned into its near monopoly status in the x86 market, and it was all downhill from there.

More specifically, the most obvious indication of Intel in decline were its troubles with the 10nm node. Originally scheduled to go into production in 2016, Intel first