#126 A Case for Open Hardware
In this edition of Technopolitik, Rijesh Panicker explains why supporting open source tools and standards is imperative. In his debut piece for Technopolitik, Tannmay Kumarr Baid argues that while recent DoD and Apple investments in MP Materials create a symbolic "mine-to-magnet" US supply chain, it remains incomplete and reliant on foreign imports due to a critical lack of domestic Dysprosium.
This newsletter is curated by Adya Madhavan.
Technopolitik: Can Open-Source Free the Chip?
— Rijesh Panicker
Over the last few months, the US has escalated its chip sanctions against China, with the latest focus being on Electronic Design Automation (EDA) software. In May, the Trump administration ordered the key EDA software providers, such as Cadence, Synopsys and Siemens EDA, which hold about 80% of the Chinese market, to stop providing services to Chinese customers. This is part of a broader strategy aimed at halting China's progress in designing cutting-edge chips. Although Chinese EDA firms have stepped into the breach and Huawei’s chairman Ren Zhenfei expressed confidence that open source software tools will help China make up the gap in their chips, for now, it seems as though the EDA chokepoint will hurt the Chinese.
This leads us to the question of open-source hardware and open-silicon. The software industry and open-source show us that it is possible to build sustainable and competitive business models. Is it possible to do the same for hardware?
Broadly, we can think of three components which together support the development of open-source hardware: 1. EDA tools: These tools automate various aspects of chip design. Since proprietary tools restrict the publication of any design output, this is a necessary condition for any open-source chip ecosystem to survive. 2. Process Design Kits (PDKs): These are design documents and libraries provided by fabrication companies (fabs) that describe the physical and electrical parameters of the basic building blocks of integrated circuits (ICs). Traditionally closed source, since 2020, several projects such as the Google sponsored SkyWater 130nm CMOS, GlobalFoundries 180nm node and the European IHP-Open130-G2 have shown the viability of open-source PDKs in driving open chip design. 3. Open-source Design IP: These represent reusable chip components, like the RISC-V processor cores, whose designs are open.
So far, there have been a few private initiatives to develop this space, such as OpenLane, which integrated DARPA’s OpenRoad EDA tool output with the SkyWater PDK, the open-source PDKs listed above and multi-wafer runs (where multiple designs are etched onto a single wafer so smaller designers can share costs) sponsored by Google, efabless and others.
In terms of government efforts in the space, DARPA’s OpenRoad EDA project was the first such effort. Ironically, DARPA’s OpenRoad EDA project was de-funded at the end of 2023 (efforts continue to get private funding for the project), after complaints from the EDA industry that it posed a serious challenge to the US’s supremacy in the space. This leaves the EU as the only region with any plan to support open-source hardware and silicon.
The EU’s open-silicon roadmap includes efforts such as sponsoring an alternative to the OpenRoad project for EDA tools, coordinating European efforts, and funding the development of open-source PDKs, starting from 130nm chips and extending down to 20nm. A key aspect of the EU roadmap is the recommendation to ensure that publicly funded tools be published with a forever-open license like the GPLv3 to prevent them from being absorbed by mainstream companies.
India’s semiconductor mission is remarkably silent on measures to promote or fund open hardware processes and projects. As my colleagues Pranay Kotasthane and Satya Sahu describe here in far greater detail, there are several opportunities to fund and support the beginnings of an open-source hardware movement. The most salient of these would be to change the focus of Semiconductor Research Laboratories (SICL) to become an open foundry supporting multi-wafer runs based on open and standardised PDKs.
Supporting open-source tools and standards for hardware is no longer optional but necessary. A robust open-source ecosystem for hardware is essential to ensure that India can maintain its autonomy in the future. It is also a model that can support and accelerate indigenous research and innovation in chip design and support India’s long-term goals in this space.
Technopolitik: The Achilles' Heel of America's Magnet Plan
— Tannmay Kumarr Baid
It’s been a good week for you if you own MP Materials stock. Two investments to bolster the company’s rare‑earth refining capabilities, one from the Department of Defense and one from Apple, have sent the shares soaring. The Pentagon is extending a $150 million loan and buying $400 million of convertible preferred equity. Apple also has hopped on the bandwagon with a $500 million magnet‑supply commitment that includes a $200 million pre‑payment for deliveries starting in 2027. In less than a fortnight MP has secured two deep‑pocketed customers and a valuable price floor: the DoD will top-up shortfall if the market price of NdPr falls below $110 per kg, roughly 75 percent above today’s spot price.
For the DoD deal, three points matter. First, routing the deal through the Pentagon, not the Commerce Department, symbolically frames rare‑earths as a national‑security asset. Second, the price floor plus a 10‑year offtake guarantee protect MP from the familiar Chinese tactic of dumping cheap oxides whenever international diversification projects gain traction. Third, direct equity makes DoD the largest shareholder; Washington now has a drivers seat inside the last U.S. rare‑earth miner.
How big is the magnet deal?
MP’s Independence campus in Fort Worth is already casting pilot magnets and has reached roughly 1,000 t per year of NdFeB (neodymium-iron-boron magnet) output as of 2024. The DoD‑financed expansion would take the total to 13,000 t per year. For context, the U.S. imported ~8,000 t of NdFeB in 2024, mostly from China. A 13,000‑t domestic run could cover all visible imports and a sliver of the magnets embedded in motors, phones, and medical devices.
However, cost optics still look out of place. Average cost for 2024 imports was roughly $40 per kg. If NdPr feedstock is held at the promised $110 per kg and every other cost matches China, finished U.S. magnets would clear around $60 per kg, just under 50 percent higher than imports. Goes to show the cost premium that the US is willing to bear to diversify the supply chain out of China.
The Dysprosium hole
The big snag is Dysprosium. Dysprosium (Dy) endows NdFeB magnets with high‑temperature coercivity; without it, traction motors and jet‑fighter actuators would roast. Typical automotive and aerospace grades use 2-3 mass-percent Dy, and even advanced processes rarely fall much below 1.5 mass-percent.
Therefore, as highlighted by the Rare-Earth Observer, at full utilisation, a headline 13,000 t of magnets, MP will need on the order of 260‑390 t of Dy metal annually. Mountain Pass simply does not have this much Dysprosium. In Mountain Pass, Dysprosium is found in Bastnaesite ore, which carries it in the tens‑of‑ppm range, translating to just about <0.05 % of total Rare Earth Oxide (REO) in the mine. Even if MP reaches its “Upstream 60K” goal (as stated in their DoD agreement) of 60,000 t Total REO production, the mine would yield barely 12 t of Dy oxide a year, under five per cent of what the magnet programme consumes.
To close the gap internally, the mine would have to raise output four‑fold to roughly 250,000 t Total REO production a year. If implemented, such a move, barring the obvious infeasibility, would also end up stripping the mine of its proven and probable reserve base in just about six years. Such a push would also unleash a glut of light REEs, lanthanum, cerium, samarium, that already face weak pricing, while NdPr oversupply would undermine the very price floor Washington just set.
Heavy‑rare‑earth processing is still a blank page
MP does not yet possess a heavy‑REE separation circuit, and its new concentrator and oxide lines are tailored to the Mountain Pass mine, which is predominantly Light REE in terms of concentration. The DoD agreement obliges the company to “extend separation activity to heavy rare‑earths.” Still, up till this point, there is no capital expenditure, flowsheet or commissioning timeline that is public. Hypothetically, even if such a plant appears, the mine’s Dy grade is so low that the unit costs of turning bastnaesite into segregated Dy oxide will be disproportionately higher over peers processing Dy internationally.
Contractually, the Pentagon can then ask MP to source feedstock from “other American suppliers.” However, the US has no permitted heavy‑REE mine, and the two planned projects are not operational (Texas Round Top is still on pilot scale, and Bear Lodge has been mothballed for almost a decade).
Thus, what will effectively happen is a shift in import reliance rather than actual full-scale supply-chain diversification. Reliance merely shifts from NdPr oxide to Dy oxide unless parallel mining breakthroughs materialise, and such breakthroughs lie outside MP’s balance sheet and the scope of the current deals.
Symbolism as the objective
What the Pentagon‑Apple double does is help a narrative of “mine‑to‑magnet made in America.” What it does not buy is a heavy rare‑earth supply chain. Dysprosium remains a missing piece of the puzzle, and the contracts overshadow it with larger, headline numbers on the NdPr front. Unless a domestic heavy‑REE source emerges, or downstream engineers abandon Dy‑bearing grades, the U.S. still imports REEs. The equity stake, price floor, and magnet orders may stabilise MP’s cash flow, but they cannot conjure Dy from a bastnaesite seam that barely contains it.