Stack Logo

2D transition metal dichalcogenide (TMD) semiconductors

The buzz around pursuing smaller, more efficient transistors has led to a deep interest in using 2D TMDs instead of Si Gate all-around (GAA) nanoribbons (NRs) below the 10nm gate length. At this tiny scale, Si faces issues like direct tunneling causing leakage, while 2D TMDs step in with their solid band gap and sustained mobilities, offering solutions to these hurdles. In the GAA NR structures, 2D TMDs showcase a scaling advantage: you can fit six 2D TMD NRs for every four Si NRs, potentially hinting at superior performance in the same space. These inherent physical benefits promise significant scaling potential.

Fabricating 2D TMD GAA NRs requires key steps like deposition, etching to form channels, doping, and contact/gate formation. However, each stage faces obstacles compared to established Si approaches. Etching TMD layers into nano-channels risks edge defects affecting performance and relies on selective chemistries that avoid harming delicate monolayers. Forming quality contacts requires optimizing exposed TMD surface area while minimizing material degradation from etching, unlike Si, conventional ion implantation doping methods damage TMDs, necessitating remote charge-based approaches that introduce coulombic scattering. Depositing effective gate oxides also struggle for a lack of dangling bonds to initiate controlled growth. Additionally, the overall mechanical fragility of atomically thin TMDs makes them vulnerable to internal stresses during processing.

Realizing the promise of 2D TMDs necessitates advancing their growth quality closer to the epitaxial precision of Si GAA NRs. Silicon’s unmatched defect densities have enabled its perseverance as the material of choice for over 50 years. Lateral growth from random nucleation sites results in polycrystalline morphologies with performance-sapping grain boundaries. Compounding this issue is the conflict between existing approaches to tackle contact resistance, primarily based on evaporated top contacts and the implicit preference for edge contacts in the simplest fabrication of NR transistors. A potential solution involves altering etch sequences to form a partial wrap-around contact, albeit potentially impacting transistor density.

Advancements in 2D TMD growth, contact resistance mitigation, and innovative fabrication techniques will likely pave the way for these materials to revolutionize the semiconductor industry, potentially rivaling or outperforming the capabilities of Si-based technology.

Apple and Arm sign deal for chip technology that goes beyond 2040

Apple has struck a deal with Arm through 2040 and “beyond,” Arm said in a U.S. Securities and Exchange Commission filing Tuesday.

The news indicates that Apple has secured access to a core piece of intellectual property, the Arm architecture, used in its iPhone and Mac chips, for the foreseeable future.

Arm, owned by SoftBank, is set to debut on the Nasdaq stock exchange in the coming weeks at a total valuation that could be as high as $52 billion, which would be the biggest technology initial public offering this year.

For Arm, its note about the Apple deal indicates that at least one of its most important partners will continue to use the company’s technology for years, quelling some fears that the change in Arm’s corporate structure could prompt some of its customers into looking for technological alternatives.

“Further, we have entered into a new long-term agreement with Apple that extends beyond 2040, continuing our longstanding relationship of collaboration with Apple and Apple’s access to the Arm architecture,” Arm said in its updated SEC filing.

Arm’s architecture is used in nearly every smartphone chip, including Apple’s A-series for iPhones. Arm’s instruction set outlines how a central processor works at its most basic level, such as how to do arithmetic or access computer memory. Switching large software projects to other instruction sets is expensive, difficult and time consuming.

Arm, originally founded in 1990, started growing fiercely after the iPhone came out in 2007 and smartphone makers needed chips that were geared for low-power usage, especially compared with the x86 architecture used in PC and server chips by Intel and AMD.

Cornerstone investors

One reason firms such as Apple use Arm’s architecture is because it has not been owned by a competitor. Arm, a British company, licensed its technology to all comers, and its customers could plan to invest billions in developing Arm chips without worrying that their access to the technology could be curtailed.

The company said 250 billion chips have shipped using Arm’s architecture, although about half the company’s royalties revenue comes from products released between 1990 and 2012, according to the filing.

Concerns over access to Arm technology is one of the main reasons regulators blocked Nvidia’s bid to buy Arm early last year, leading to this fall’s IPO.

Apple, GoogleNvidia, Samsung, AMDIntelCadence, Synopsis, Samsung and Taiwan Semiconductor Manufacturing Company have expressed interest in buying some Arm shares as part of the offering, as much as $735 million in total according to the filing, which would give those companies a stake in Arm’s ownership and some say in how it is managed. They’re referred to as “cornerstone investors.”

China Scuttles a $5.4 Billion Microchip Deal Led by U.S. Giant Intel

Intel calls off its planned acquisition of Tower Semiconductor, an Israeli chip maker, after waiting in vain for 18 months for a review by Chinese regulators.

China has effectively scuttled a $5.4 billion deal by Intel, the Silicon Valley semiconductor giant, in the latest sign of the frayed business ties between China and the United States.

Intel, which has long had operations in China, said Wednesday that it had “mutually agreed” to terminate a planned merger with Tower Semiconductor, an Israeli chip manufacturer. The announcement came after China’s antitrust regulators failed to rule on the transaction before a deadline set by the companies.

The failure of Intel to complete the acquisition of Tower could send a further chill through American companies with deep ties in China, where it is becoming increasingly difficult to do business amid tensions between the two countries.

The planned merger, announced in February 2022, passed an antitrust review in the United States and several other geographies. But it ran into a lengthy delay in China, where regulators review mergers of companies that earn a certain amount of revenue in the country.

Technology is the prime battlefield in the tense economic relations between China and the United States.

Beijing is deeply upset by an American-led set of international restrictions on the sale to China of the most advanced computer chips, which have military applications, and of the factory equipment to make such chips. Those restrictions were put in place in October. In a separate action, President Biden last week ordered a ban on certain new investments in sensitive Chinese technology.

China has condemned the moves as an effort by Washington to throttle its tech development and slow its economic growth.

Despite the raw tensions between the countries, their economies remain highly interconnected, dependent on one another’s supply chains, technology and investment money.

For Intel, China is both a major marketplace and place of business: In 2022, the company employed more than 12,000 people there, and made more than $17 billion in revenue, about 27 percent of its global total. It started doing business in China in the mid-1980s, with operations that include assembling and testing chips manufactured elsewhere.

Intel, which is struggling to regain a lead in chip production technology, hoped the merger with Tower would help accelerate a shift to become a major manufacturer for other designers of chips. Intel has previously mainly used its factories to produce chips it both designs and sells.

Tower, which has an office in Shanghai, was founded in 1993 and operates a relatively small chip manufacturing service compared with giants like Taiwan Semiconductor Manufacturing Company. Intel will pay Tower $353 million for failing to close the deal, according to a statement by Intel.

Intel’s inability to get the merger approved in China underlines what could become an increasingly hard choice for multinationals: They may need to choose between having operations in China or carrying out mergers and acquisitions around the globe. Such concerns could produce a further chill on foreign investment in China, which has already plunged this year because of geopolitical concerns.

The Chinese government agency that decides whether to approve global mergers, the State Administration for Market Regulation, is now “in an uncomfortable spotlight as a proxy for China’s commitment to market access for foreign investors,” said Han Shen Lin, the China country director for The Asia Group, an advisory firm in Washington.

Before the agency was established in 2018, global mergers were reviewed in China mainly by a unit of the Ministry of Commerce, which is dominated by civil servants with extensive international experience and contact with foreign businesses and governments.

The State Administration for Market Regulation, by contrast, is categorized within the Chinese bureaucracy as primarily a domestic agency, and its officials have shunned most contact with foreign governments, embassies or businesses.

Patrick Gelsinger, who became Intel’s chief executive in early 2021, has pushed to add what the industry calls chip foundry services, in part to attract U.S. government subsidies under legislation passed a year ago. He recently traveled to China to help get the Tower deal approved.

“We continue to drive forward on all facets of our strategy,” Mr. Gelsinger said in a statement on Wednesday.

Intel’s fabrication plants, or fabs, tend to specialize in advanced production processes used to make microprocessors and other digital chips. Tower, by contrast, is best-known for older technology that produces analog chips, which are used for jobs like amplifying signals and managing power in cellphones and other products.

The company now owns two fabs in Israel, two in the United States, three in Japan and is participating in a joint manufacturing venture in Italy.

Seven major automakers are teaming up on a North American EV charging network

BMW, GM, Honda, Hyundai, Kia, Mercedes-Benz, and Stellantis are creating a joint venture to launch a public charging network in North America.

A new group of automotive super friends is banding together, promising to build the next big North American electric vehicle charging network. These worldwide automakers — BMW, General Motors, Honda, Hyundai, Kia, Mercedes-Benz, and Stellantis — announced a planned joint venture today to erect easy-to-activate DC fast chargers along US and Canadian highways and in urban environments.

The grand plan for the currently unnamed partnership is to install “at least” 30,000 high-speed EV chargers by 2030, with the first ones to open summer 2024 in the US. The collective plans to leverage National Electric Vehicle Infrastructure (NEVI) funding in the US and will also use other private and public funding from state and federal sources to build out the network.

Current EV charging networks, from Tesla Superchargers to Electrify America, have stations installed in places where people can shop, eat, and use the bathroom. In a similar fashion, these new chargers will also be installed along routes to vacation destinations and in metropolitan areas.

The new stations will connect and charge EV models made by the partnered automakers without having to fumble with another charging station app. The companies also plan to integrate the developing “Plug and Charge” standard that the Federal Highway Administration is attempting to standardize.

“The better experience people have, the faster EV adoption will grow,” GM CEO Mary Barra states in the joint venture press release. Many charging networks today require their own apps and have issues of reliability. Tesla’s Supercharger network, which is considered among the best in the world, will be available to vehicles from automakers including FordGM, Volvo, and more without needing people to activate with an app.

The new joint venture is also planned to be entirely powered by renewable energy. It’s not known if renewable energy will directly power them or if the companies plan to buy credits like Rivian announced yesterday.

Canadians will have to wait for “a later stage” before initial stations are installed. All stations will include the standardized Tesla North American Charging Standard (NACS) ports and also the current widely used Combined Charging System (CCS) plugs.

The Boring Future of Generative AI

ChatGPT’s chaotic streak can be charming. Google’s new chat-style search shows text-generation technology is headed in a much tamer direction

THIS WEEK, AT its annual I/O developer conference in Mountain View, Google showcased a head-spinning number of projects and products powered by or enhanced by AI. They included a new-and-improved version of its chatbot Bard, tools to help you write emails and documents or manipulate images, devices with AI baked in, and a chatbot-like experimental version of Google search. For a full recap of the event, complete with insightful and witty commentary from my WIRED colleagues, check out our Google I/O liveblog

Google’s big pivot is, of course, largely fueled not by algorithms but by generative AI FOMO. The appearance last November of ChatGPT—the remarkably clever but still rather flawed chatbot from OpenAI—combined with Microsoft adding the technology to its search engine Bing a few months later, triggered something of a panic at Google. ChatGPT proved wildly popular with users, demonstrating new ways to serve up information that threatened Google’s vice grip on the search business and its reputation as the leader in AI

The capabilities of ChatGPT and AI language algorithms like those powering it are so striking that some experts, including Geoffrey Hinton, a pioneering researcher who recently left Google, have felt compelled to warn that we might be building systems that we will someday struggle to control. OpenAI’s chatbot is often astonishingly good at generating coherent text on a given subject, summarizing information from the web, and even answering extremely tricky questions that require expert knowledge.

And yet, unfettered AI language models are also silver-tongued agents of chaos. They will gladly fabricate facts, express unpleasant biases, and say unpleasant or disturbing things with the right prompting. Microsoft was forced to limit the capabilities of Bing chat shortly after launch to avoid such embarrassing misbehavior, in part because its bot divulged its secret codename—Sydney—and accused a New York Times columnist of not loving his spouse.

Google worked hard to tone down the chaotic streak of text-generation technology as it prepared the experimental search feature announced yesterday that responds to search queries with chat-style answers synthesizing information from across the web.

Google’s smarter version of search is impressively narrow-minded, refusing to use the first person or talk about its thoughts or feelings. It completely avoids topics that might be considered risky, refusing to dispense medical advice or offer answers on potentially controversial topics such as US politics.

Google deserves recognition for reining in generative chatbots’ wild side like that. But in my tests, the new search interface felt incredibly tame compared to ChatGPT or Google’s own chatbot Bard.

As the company moves the technology into more of its products, perhaps the generative AI revolution will turn out to be a lot less fun than you might expect from the early shock and awe of ChatGPT, a chatbot that has an edgy charm. Gone are the wild ravings and imaginings of powerful AI bots. In their place are new ways to populate spreadsheets, compose email pleasantries, and find products to buy.

Even if “AI doomers” warning about errant AI prove overblown, it will be interesting to watch how companies like Google and OpenAI balance the development of more powerful generative language models with the need to have them behave.

Google has invested huge sums and major resources in AI over recent years, with CEO Sundar Pichai often pitching the company as “AI first”, and the company is desperate to show it can advance the technology more quickly than OpenAI. One high-level message from Google’s stream of AI announcements was that the company is not going to hold back anymore, as it did the LaMDA chatbot that was announced long before ChatGPT appeared but not made public.

In March, some big names in AI research signed an open letter calling for a six-month pause on creating machine learning systems more powerful than GPT-4, which powers ChatGPT. Pichai was not a signatory and said in his keynote speech yesterday that the company is currently training a new, more powerful language model called Gemini.

A source at Google tells me this new system will incorporate a range of recent advances from different large language models and may eclipse GPT-4. But don’t expect to get to experience the full power or charisma Gemini can offer. If Google applies the same chaos-taming methods seen in its chat-like search experiment, it may just seem like another surprisingly clever autocomplete.