After 30 years of holding the crown as the world’s most valuable semiconductor company, Intel is selling below book value. At today’s price, Intel is an affordable acquisition for many, even much smaller tech companies. But what happens to Intel’s factories, designers, and intellectual property is vitally important.

Intel is the only large-scale American manufacturer of advanced logic semiconductors, even if it is no longer leading at the cutting edge. During the pandemic, we learned that shortages of semiconductors can have a devastating impact on the economy. Moreover, we need to develop leading-edge capacity to stay ahead in many advanced computing and defense-related technologies, including artificial intelligence. Most potential buyers of Intel would likely focus on cost-cutting and see little value in a money-losing manufacturing subsidiary (known in the industry as a foundry). In other words, profit-seeking buyers cannot be depended upon to maintain America’s manufacturing capability.

The entire world benefits if Intel has world-class capacity. TSMC and Intel have been competing for chip leadership for 30 years. Until seven or eight years ago, Intel was winning this battle. While this never-ending competition has made the world richer and must continue, depending on a single manufacturer strategically located with its most advanced factories in Taiwan is a global risk. In semiconductors,  process research and development and leading-edge manufacturing must be co-located. So TSMC will never move its most advanced technology to the U.S.  It is simply too expensive, requires too much infrastructure and too many key employees would have to move. The U.S. needs Intel.

Intel’s management and board must take responsibility and move decisively to stem the bleeding. The actions they’ve taken to date will not suffice. Numerous Intel watchers (including ourselves) have argued for several years that it must separate its foundry/manufacturing business from its design business. An Intel foundry operation, inside Intel’s corporate structure, has little chance of success. 

Nvidia, Qualcomm, Broadcom, and others are desperate for a second manufacturing option to TSMC, but will remain hesitant as long as Intel directly competes with them. Samsung, the only other advanced manufacturer of chips, has similarly discovered that many chip designers such as Apple and Nvidia tend to avoid its foundry because Samsung is a potential competitor. Intel’s management has also failed to prove that it can effectively run a foundry.  Intel offered foundry services to the industry for two decades, never building a successful business. Missed targets and deadlines and management turnover do not inspire confidence.  

Intel’s CEO, Pat Gelsinger, is a true technologist who played an important role in the company’s storied past. Today, he’s faced with a difficult decision: whether to break up the iconic company. He already announced a plan to establish Intel Foundry as an independent subsidiary inside Intel. But this doesn’t go far enough. Emotions aside, the path for the U.S. and Intel should be clear.

Since Intel’s design business remains profitable, it would need to establish a long-term supply contract with a newly created foundry independent of Intel. Just as AMD separated its manufacturing in 2009 and launched Global Foundries with a long-term supply deal, the new Intel design company would need to partially underwrite the foundry’s losses and guarantee sales for several years.

Intel’s design company alone cannot support an independent foundry. Yet, Intel’s manufacturing operation is the only hope for maintaining the most advanced nodes on U.S. soil. An independent foundry would offer open access to all American, Korean, Japanese, and European companies to accrete sufficient volume and ensure its commercial viability.

Since this is a public good (all of Intel’s current competitors and customers, as well as U.S. and global consumers, would benefit), the U.S. government (in cooperation with allies) can and should play a pivotal role.

The CHIPS Act gives the U.S. government $39 billion in grants to revive American semiconductor manufacturing. The government has already promised (but not yet disbursed) up to $8.5 billion in grants and $11.5 billion in low-cost loans for Intel. Today, Intel threatens to become this administration’s Solyndra (the solar company, which went bankrupt after getting more than $500m in government funding). This would be disastrous, both for the government and Intel. The government has the leverage to force Intel down a better path—and it must use it now.

You cannot have an industrial policy without an industry: The government should be very clear on what it is willing to finance, including Intel’s corporate structure. This means that the government should insist that design and manufacturing at Intel be severed into two truly independent companies.

Time is not on our side. It took Intel less than a decade to lose its lead—and it will take at least five years to get back in the game. The pace of change in the chip industry demands quick action by management and the government. TSMC is not slowing down. The longer we wait, the less competitive we become.

More must-read commentary published by Fortune:

• COVID-19 raises the stakes for heart attacks, strokes, and even death long after infection, new study finds
• Here’s why the gap between Americans’ perception of U.S. economic performance and reality has doubled since 2019, according to economists
• The U.S., EU, and India must urgently come together in a free trade agreement on clean energy, says the CEO of one of India’s largest renewables companies
• Here are the facts that activist critics of Pfizer CEO Albert Bourla are missing, according to a Yale analysis

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.