Few brands in technology carry the storied pedigree of Intel. For decades, “Intel Inside” was synonymous with progress in computing—an American icon that repeatedly raised the bar for processor performance and innovation. Today, Intel finds itself in a precarious position. Its once-vaunted manufacturing machine has fallen behind TSMC, which dominates advanced process nodes. Meanwhile, Intel’s core CPU design business—a substantial, if bruised, revenue engine—remains overshadowed by the mounting losses and capital burn in its foundry operations. Rumors swirl that Broadcom might step in and scoop up Intel’s design arm. But for those who believe in competition, national security, and the long-term integrity of an American technology brand, that outcome would be deeply misguided.
In this Special Breaking Analysis, we examine why Intel’s design business can—and should—remain an independent powerhouse once the company spins out its foundry operations. We’ll walk through how a spinout might unfold, including an updated scenario from our original bold plan in which TSMC anchors a U.S.-domiciled joint venture. We’ll also detail why the foundry must command a massive discount—perhaps as deep as 90–100% of Intel’s sunk investment—and how a “take-or-pay” wafer pre-purchase model from hyperscalers (Amazon, Google, Microsoft, Apple, Meta, and NVIDIA) can sustain a new American-based foundry while freeing Intel to focus on design excellence.
Along the way, we’ll explore the role of advanced nodes like Intel’s much-touted 18A, dissect why we feel the assumptions of IP value around 18A might not be as viable as the pundits say. We hope to convey why all these moves serve a broader purpose: to maintain robust competition, fortify national security, and preserve an American chip design champion.
Intel’s Current Predicament
Intel’s integrated manufacturing strategy was once a key differentiator. By aligning design and process technology in tandem, Intel could produce new microprocessors at a pace the rest of the industry struggled to match. But TSMC’s rise—fueled by enormous wafer volumes from Arm-based mobile processors and an unwavering focus on manufacturing excellence—has left Intel behind in advanced process nodes. The capital intensity of staying competitive at the leading edge has skyrocketed, and Intel’s foundry efforts have yet to attract sufficient scale or external customers.
Pat Gelsinger’s return as CEO initially reignited optimism for a U.S.-based advanced manufacturing renaissance. But the delays in Intel’s roadmap, the cost overruns in ramping new nodes, and the persistence of TSMC’s leadership suggest that Intel’s foundry plan remains outgunned. Meanwhile, the design side of Intel’s house—responsible for everything from x86 CPUs for PCs and servers to emerging AI accelerators—still has enormous potential. The problem is that this potential remains shackled by the continuous drain of Intel’s foundry ambitions.
For more context, read our Critical Analysis on the Root Cause of Intel’s Troubles.
Why Selling the Design Arm to Broadcom or Another Suitor Is a Mistake
Rumors of Broadcom acquiring Intel’s design business reflect a short-term mindset. Broadcom has built its empire through acquisition, focusing on cost efficiencies and leveraging synergies across enterprise software, networking, and semiconductors. Intel in many ways would fit Broadcom’s strategy of acquiring durable businesses. Handing over Intel’s CPU design business—a bedrock of American microprocessor leadership—would inevitably concentrate more market power. That’s not only a threat to healthy competition; it risks overshadowing Intel’s legacy and future innovation. Given the heightened scrutiny on tech M&A by both U.S. and international regulators, a Broadcom-Intel design merger could and should face drawn-out investigations or outright rejection.
Moreover, Intel’s stock price is depressed compared to historical peaks, in no small part because of the drag from manufacturing. Why sell the design arm now, when it’s arguably at a cyclical low? Once the foundry baggage is removed and Intel’s design group can stand on its own merits—much like AMD did when it spun off GlobalFoundries—there’s every possibility for a resurgence. Selling at the bottom of the cycle would short-change shareholders and kill an iconic brand that still has real technological capital.
The Board’s Responsibility to Preserve an American Icon
Intel’s board owes it to shareholders, employees, and the broader tech industry to keep the design business independent. Yes, the integrated manufacturing approach has proven costly, and yes, the foundry is likely a negative-value asset as it stands. But there is precedent for this kind of strategic realignment: IBM paid GlobalFoundries to take its microelectronics division in 2014. AMD separated its fabs in 2009 and eventually became a formidable design-centric competitor. Today, AMD relies on TSMC to manufacture its designs at leading-edge nodes. With the money drain from manufacturing gone, AMD could focus on architectural innovation—an effort that has paid off handsomely.
In the same vein, Intel can rebuild its design brand and engineering culture once it is liberated from the massive overhead of running advanced-node fabs. The board shouldn’t tap out and sell the design side to a competitor; it should step up and ensure that an iconic American brand continues to innovate in CPU, GPU, AI, and custom silicon markets.
Spinning Out Intel’s Foundry and Creating the “Foundry of America”
Negative-Value Assets and the IBM Precedent
Intel in our view has no choice but to exit manufacturing. To spin foundry will likely require a substantial discount—some might even call it “paying someone to take it off their hands.” In 2014, IBM effectively wrote GlobalFoundries a large check to absorb its microelectronics business, transferring factories, employees, and IP. Intel’s foundry operations have consumed billions in capex with underwhelming yields and missed process milestones. Critics may point to 18A as a potential game-changer, but the reality is that 18A is behind schedule, unproven, and not generating revenue.
To entice a partner like TSMC to operate these fabs in the U.S., Intel must accept that much of its capital investment will not be recovered. A 90–100% discount relative to the book value of Intel’s foundry assets may sound extreme, yet it’s not unprecedented when the alternative is ongoing, multi-billion-dollar losses. This is the cost of a strategic reset.
Why 14A Could Matter More Than 18A
Intel has touted 18A as the company’s hope to leapfrog TSMC. But each roadmap slip erodes market confidence and 18A’s negative profitability was never going to allow Intel to catch TSMC with this node . The deeper truth is that the real wave of future manufacturing innovation was more likely with what Intel calls 14A (or some iteration of that node)—combining backside power delivery, gate-all-around, and next-generation High-NA EUV in a more mature process window. We’ve always said this was an extremely risky and unproven endeavor but in the unlikely event Intel could pull it off, it would have likely represented a leapfrog in process relative to TSMC. But the probability of success in a reasonable timeframe was always very low.
However, 14A in the JV’s hands under TSMC’s guidance would likely be more fruitful. Because it has profitable advanced nodes already, it could take its time with 14A rather than trying to catch up in desperation. TSMC has the financial and operational discipline to properly integrate new techniques. If TSMC is part of a U.S.-domiciled JV, it can exercise that discipline on what remains of Intel’s advanced node R&D while leveraging some of Intel’s IP. In this scenario, 14A would find a more reliable steward in TSMC than Intel’s own cash-strapped and time-pressured foundry operation.
A U.S.-Domiciled TSMC JV
The likely scenario is a joint venture—wholly based in the United States—where TSMC contributes expertise, IP, and skilled engineering teams deployed stateside. Unlike typical equity investments, TSMC might not contribute much upfront cash. Instead, it would bring intangible but invaluable process know-how, yield ramp experience, and world-class manufacturing talent. The U.S. government’s CHIPS Act funds could partially offset capital expenses, while private equity players provide debt financing backstopped by either Washington or the JV’s assets.
Under this arrangement, Intel’s foundry assets and workforce transfer to the JV. TSMC becomes the majority stakeholder or at least the primary operator, ensuring that day-to-day manufacturing decisions align with TSMC’s methodology. Intel, for its part, might retain a minority stake but cede operational control—effectively paying to hand over the reins to the world’s most competent foundry operator. The discount on Intel’s assets is baked into the deal; it reflects the expensive truth that Intel’s fabs have negative economic value unless someone can turn them around.
Below is a revised and simplified version of our original CAP table:
Hypothetical JV Ownership
The Take-or-Pay Model: Locking in Demand from Cloud & Internet Giants
Why the Hyperscalers Are Key
One of the greatest hurdles for a newly spun-out foundry is securing enough volume to achieve economies of scale. TSMC thrives on huge wafer starts from Apple, MediaTek, AMD, NVIDIA, and many others. Intel’s x86 business alone won’t provide the necessary volume, especially if it’s behind TSMC’s own leading-edge processes. So, the solution is to bring in the U.S. hyperscalers—Amazon (AWS), Google, Microsoft, Apple, Meta, NVIDIA, the US military and others—who each have a vested interest in:
- A second source for advanced node capacity (hedging geopolitical risk in Taiwan).
- A local supply chain for AI, HPC, and specialized ASICs.
- More favorable regulatory scrutiny in exchange for supporting a national semiconductor infrastructure.
By structuring multi-year “take-or-pay” wafer pre-purchase contracts, the hyperscalers guarantee consistent demand regardless of yield ramp challenges. This arrangement underwrites the JV’s operational costs during the toughest early years. The “take-or-pay” concept means that if yields remain low or the node is behind TSMC’s mainstream lines for a while, these cloud giants still pay for the capacity. In return, they get guaranteed supply when (and if) the JV’s yields improve, plus the political goodwill of being seen as partners in securing U.S. semiconductor independence.
Timeline for the Take-or-Pay Commitments
The crux of this model is that the guaranteed demand from hyperscalers offsets the typical scale disadvantage a new foundry would face. This approach also dovetails with the government’s desire to see advanced node manufacturing thrive on U.S. soil, thereby reducing national-security risks tied to any instability in Asia.
Big Picture: Why the Hyperscalers Participate
The internet giants—Apple, Amazon (AWS), Google, Microsoft, Meta, and NVIDIA—are placing large, multi-year bets on American silicon. But it’s not a classic equity investment. Instead, each of these firms would be committing to substantial “take-or-pay” wafer agreements, locking in critical supply for advanced semiconductors without the long wait for a traditional equity ROI. In exchange, the US government would take its foot off the necks of these companies in the form of relaxed regulatory pressure and less onerous oversight.
Guaranteed Supply on US Soil
For years, Apple and the hyperscalers have relied almost exclusively on TSMC’s Taiwanese fabs. That’s understandable—TSMC is the gold standard in advanced node manufacturing. But rising geopolitical tensions underscore the risk of tying all capacity to a single region. By participating in a new U.S.-based JV, these companies gain a dependable second source for high-performance chips, hedging against sudden disruptions.
From iPhone SoCs to AI accelerators, every hyperscaler wants an uninterrupted flow of silicon. Being first to market with new AI/ML capabilities or next-gen consumer devices can mean the difference between industry leadership and playing catch-up. A domestic advanced node foundry, even if initially lagging TSMC’s bleeding edge by a half-step, gives these firms a crucial foothold in American-made chips—a prospect that’s become priceless.
US Government Eases Regulatory Pressure
What do these hyperscalers get beyond supply assurance? A more predictable regulatory climate. In Washington, the appetite to scrutinize Big Tech—whether over antitrust, privacy, or monopolistic practices—has only grown. Committing massive wafer prepayments to a strategic U.S. foundry aligns their corporate goals with the government’s national interests.
By no means are we suggesting these firms get a total regulatory pass. But the sheer scale of their take-or-pay deals supports America’s push for domestic semiconductor leadership, buying goodwill and tempering how aggressively regulators wield the antitrust hammer. For these companies, predictability in the legal arena translates into billions of dollars in strategic value. Moreover, the US government in our scenario would pressure EU regulators to back off as well, in exchange for a more stable supply chain for advanced silicon. Since the US is funding the future of silicon, it has the moral authority to pressure EU regulators to moderate their anti big tech stance against US companies.
Instant Impact vs. a 10-Year Wait
By opting for wafer guarantees rather than equity stakes, the hyperscalers sidestep the classic venture capital pitfall of waiting a decade for payback. Every one of these companies is in constant need of leading-edge silicon; paying upfront for guaranteed capacity isn’t just a hedge—it’s an essential input to their day-to-day business. So unlike a typical long-horizon investment, these wafer commitments deliver near-term strategic value: a stable supply of advanced chips to fuel growth in AI, cloud services, consumer devices, and more.
In return, the new foundry scores sustained demand. Even if early yields lag behind TSMC’s global best, the arrangement smooths out the revenue cycle while the JV works to improve yield and process maturity. It’s a symbiotic setup: Big Tech gets supply; the foundry gets a stable customer base and a realistic shot at scale.
Snapshot of the Hyperscaler Motivations
Apple
Apple has long been TSMC’s shining star, but a single point of manufacturing failure is no longer tenable. Diversifying to a domestic foundry paves the way for second-source security across iPhone, iPad, and Mac chips. Over the long haul, Apple might also co-develop next-gen packaging lines at the JV, ensuring it retains a competitive edge in device performance.
AWS
AWS is hungry for capacity for its Graviton CPUs, Trainium, and Inferentia AI chips, plus whatever HPC solutions loom on the horizon. And there’s a regulatory carrot: if Amazon helps secure U.S. semiconductor leadership, the government will commit to soften scrutiny around its e-commerce dominance.
Google Cloud’s TPU programs demand advanced silicon to power AI/ML workloads at scale. A domestic foundry removes the uncertainty of Taiwanese supply lines and the US government would agree to lessen its scrutiny around data privacy or large-scale acquisitions. Predictability here is priceless for a company whose approach is “move fast and innovate.”
Microsoft
Microsoft’s diverse product suite spans cloud hardware, AI chips (Project Athena), and even Xbox SoCs. A stake in U.S.-based capacity can deflect the usual antitrust chatter around Office and Windows bundling. The real boon is ensuring timely access to advanced nodes for its AI and console businesses.
Meta
Meta faces heavy scrutiny on social media policy and data privacy. Securing advanced-node chips for AI inference, training, and upcoming AR/VR headsets is mission-critical. Making big wafer commitments demonstrates goodwill that could soften regulators’ stances on everything from user data to content moderation.
NVIDIA
As the go-to provider of GPUs for AI and gaming, NVIDIA constantly grapples with supply shortages—and no one knows that better than data center customers. Having a domestic fab option helps shield against overseas risk and further ingratiates NVIDIA to U.S. policymakers, potentially smoothing the road for future expansions or acquisitions.
Additional stakeholders in this model could include the US military, Qualcomm, Tesla, MediaTek, AMD and possibly IBM, which could also contribute IP from its advanced research in chips.
Bottom Line: Strategic Payoff
A U.S.-based advanced foundry JV, underpinned by these massive “take-or-pay” agreements, is more than a financial arrangement. It’s an ecosystem-level deal that ties the fortunes of Big Tech to America’s semiconductor resurgence. Each hyperscaler hedges against global uncertainty and gains a modicum of regulatory relief in the process. Meanwhile, the JV secures guaranteed demand from day one—enough to build momentum toward competitive yields, node maturity, and long-term viability.
In short, this is a bet on stability—geopolitical, operational, and regulatory. And at a time when advanced silicon has become the oil of the digital economy. We believe that’s an investment well worth making.
Hypothetical Take-or-Pay Commitments
Total Upfront Payments: $25B
Total Milestone Payments: $19B
Grand Total: $44B (over 5 years)
Note: These figures are illustrative to show a hypothetical scale. They would be adjusted based on overall fab construction costs, negotiated TSMC licensing fees, and a reasonable node volume.
This structure, when combined with CHIPS Act funding and debt financing, gives the new American Foundry approximately $100B in capital over a five year period. With TSMC’s leadership, process expertise and talent injection it dramatically increases the chances of having a viable advanced semiconductor manufacturing source on US-shores with a US domiciled company.
Intel as an Independent Design Powerhouse
Reinventing “Intel Inside”
Once the foundry is spun out, Intel can refocus on its core: designing innovative CPUs, GPUs, AI accelerators, and custom silicon solutions. By removing the financial and managerial overhead of manufacturing, Intel’s leadership can redeploy cash toward R&D in architecture, software stacks, packaging innovation, and go-to-market partnerships. A streamlined Intel might look a lot like AMD, NVIDIA, or Qualcomm—fabless (or fab-lite), yet able to bring advanced chip designs to market quickly by tapping TSMC’s best processes.
Intel could still have preferential access to the newly created JV, securing capacity for its own x86 lines at better rates or guaranteed volumes. But if that JV lags TSMC’s mainline by too much, Intel’s design arm could also choose to go direct to TSMC’s Taiwan fabs or Samsung’s foundries. This freedom would be an enormous boost to Intel’s competitiveness. No longer chained to a sub-scale manufacturing model, Intel design engineers can choose whichever advanced node best suits the product timeline and competitive environment.
Preserving Competition and U.S. Leadership
An independent Intel design division also satisfies the concerns of regulators who fear large-scale consolidation. Selling the design arm to Broadcom would add to the wave of semiconductor M&A that the government is already carefully scrutinizing. By staying independent, Intel remains a distinct competitor in CPU, HPC, and AI markets. The U.S. Department of Defense also gains because Intel can keep focusing on specialized designs for defense-related applications, while the new JV foundry (with TSMC stewardship) handles secure manufacturing on American soil. It’s a synergy that retains critical R&D in the United States, fosters multiple points of competition, and ensures a stable supply chain for military and civilian infrastructure.
Addressing Skeptics and Risk Factors
1. Will TSMC Truly Commit to the U.S.?
TSMC has already broken ground on fabs in Arizona, though it has struggled with costs and workforce challenges. Critics question whether TSMC will allocate its best technology to U.S. sites. In the proposed JV scenario, TSMC could opt to transfer a slightly older node to ensure it retains absolute leadership in Taiwan. The new entity might operate at “N-1” levels of technology. Yet for many hyperscaler workloads—like server CPUs, AI inference chips, or specialized accelerators—being a generation behind might be acceptable when weighed against the security of local supply. Over a decade, with guaranteed volume and financial stability, that JV could close some of the gap if TSMC invests in incremental process improvements.
2. Might the Government Resist a TSMC JV?
One could argue that the U.S. government wants Intel to succeed as a standalone integrated manufacturer. But the repeated delays and huge losses reveal that Intel won’t ever catch up to TSMC alone (without unlimited capital and time). The CHIPS Act aims to ensure advanced node capacity exists in the U.S., not necessarily that Intel remains integrated if that model is doomed to fail. A TSMC-led JV, operating on U.S. soil, employing tens of thousands of Americans, and financed partly by Big Tech’s wafer prepayments, meets the spirit of the CHIPS Act. For the government, it’s a pragmatic solution to ensure advanced manufacturing doesn’t vanish entirely from American shores.
3. Can Intel’s Design Unit Survive Without a Guaranteed Fab?
A valid concern is that Intel might lose its “co-optimization” advantage by spinning out manufacturing. But the success of AMD, NVIDIA, and Apple’s silicon team indicates that close partnerships with TSMC can achieve better results than a sub-scale in-house approach. Intel’s advanced packaging technologies (like EMIB, Foveros) can be licensed to or co-developed with TSMC under commercial agreements. As an independent design entity, Intel can become more nimble, not less.
4. What if Hyperscalers Find the Yields Unacceptable?
Take-or-pay contracts assume that advanced node yields will improve over time. Early yields might be poor. The hyperscalers, however, get two things they value: guaranteed capacity (even if it’s behind TSMC’s leading edge) and better regulatory standing with the government. Over the long haul, if yields are too low and the JV fails, the hyperscalers would still have TSMC’s mainline in Taiwan (or Samsung) as a fallback. Because these are multi-year deals, the JV has enough runway to fix process issues, guided by TSMC’s real-world experience.
The Path Forward: A Vision for Intel and the U.S. Tech Ecosystem
Rather than ceding its legacy to Broadcom or any other suitor, Intel should embrace a strategic pivot that upholds competition, national security, and the spirit of American innovation. The steps are clear, albeit not easy:
- Spin Out the Foundry: Intel’s board must recognize the foundry’s negative value and take the gift proposed in our revised CAP table. While the discount might be monumental—90–100% of Intel’s sunk investment, it sets the stage for future success.
- Form a U.S.-Domiciled JV: TSMC brings IP, expertise, and key personnel stateside. Intel’s foundry assets transfer to the JV, with potential partial ownership but not operational control. CHIPS Act funds, private equity debt, and hyperscaler prepayments finance the new venture.
- Leverage Take-or-Pay: The hyperscalers (Amazon, Google, Microsoft, Meta, Apple, and perhaps others like NVIDIA or Broadcom itself) commit billions in wafer prepayments. They secure advanced node capacity on American soil, along with potential regulatory goodwill.
- Intel Design Stands Independent: Freed from the foundry’s capital demands, Intel focuses on x86 CPU leadership, discrete GPU technology, AI accelerators, and custom silicon. It can choose whichever foundry suits its roadmap—be that the new JV, TSMC’s mainline, or even Samsung—much like other fabless giants.
- Regulators Endorse the Spinout: Rather than fight a losing battle to keep Intel integrated, the U.S. government encourages the JV as a pragmatic path to onshore advanced node capability. In parallel, it blocks or discourages large acquisitions that reduce competition (e.g., a Broadcom-Intel design merger).
- A New Era of U.S. Semiconductor Innovation: Over the next decade, the JV’s yields improve, and it climbs closer to TSMC’s leading edge. Intel’s design business reclaims its creative edge, funneling R&D into architectural breakthroughs unencumbered by manufacturing constraints. The hyperscalers get consistent, if slightly lagging, advanced node supply in the U.S., and the government sees a major piece of critical infrastructure secured on American soil.
Final Thoughts
Intel’s board is under pressure to deliver shareholder value in the wake of ongoing foundry failures. Selling Intel’s design business to Broadcom might look like a quick fix, but it would sacrifice a venerable American brand at the lowest point of its valuation—an outcome that undermines competition, national security, and the legacy of one of tech’s greatest innovators. We believe it is incumbent on this board, which oversaw the flawed IDM 2.0 strategy, to preserve Intel as an independent entity and support the hard work it will take to give rebirth to this American icon.
The visionary and entirely viable route is to spin out the foundry, swallow the painful discount, and let Intel’s design house flourish independently. We’ve seen the positive results of such a strategy with AMD, which soared after distancing itself from foundry headaches. With a TSMC-led joint venture offering a U.S.-based alternative to Taiwan’s facilities—and hyperscalers incentivized to prepay for wafer volume—this scenario provides a lifeline for advanced semiconductor manufacturing in America. It keeps Intel relevant in the CPU, GPU, and AI arenas. It aligns with government aims for robust competition and supply chain security. And it ensures that “Intel Inside” remains more than just a memory.
In short, Intel’s next chapter need not be defined by a desperate sale to a conglomerate. The foundry spinout is a bitter pill, but once swallowed, it could liberate Intel’s design prowess for decades to come—reestablishing it as a powerhouse of American innovation, sustaining competition across the CPU and AI markets, and serving as a keystone in the broader effort to restore domestic semiconductor manufacturing leadership. It’s a vision that demands courage from Intel’s board, patience from its shareholders, and cooperation from the US government and partners like TSMC and the hyperscalers—but one that could ultimately salvage Intel’s future as an iconic force in technology.
