Why Google's custom AI chips are shaking up the tech industry Ironwood is Google's latest tensor processing unit Nvidia's position as the dominant supplier of AI chips may be under threat from a specialised chip pioneered by Google, with reports suggesting companies like Meta and Anthropic are looking to spend billions on Google's tensor processing units. The success of the artificial intelligence industry has been in large part based on graphical processing units (GPUs), a kind of computer chip that can perform many parallel calculations at the same time, rather than one after the other like the computer processing units (CPUs) that power most computers. 'Flashes of brilliance and frustration': I let an AI agent run my day GPUs were originally developed to assist with computer graphics, as the name suggests, and gaming. "If I have a lot of pixels in a space and I need to do a rotation of this to calculate a new camera view, this is an operation that can be done in parallel, for many different pixels," says Francesco Conti at the University of Bologna in Italy. This ability to do calculations in parallel happened to be useful for training and running AI models, which often use calculations involving vast grids of numbers performed at the same time, called matrix multiplication.

Artificial intelligence can appear to be many different things--a whole host of programs with seemingly little common ground. Sometimes AI is a conversation partner, an illustrator, a math tutor, a facial-recognition tool. But in every incarnation, it is always, always a machine, demanding almost unfathomable amounts of data and energy to function. AI systems such as ChatGPT operate out of buildings stuffed with silicon computer chips. To build bigger machines--as Microsoft, Google, Meta, Amazon, and other tech companies would like to do--you need more resources.

Elon Musk has announced that Tesla's new custom AI chip is about six months away from being installed in new production cars. The CEO said that the chip, which was confirmed as being in development last December, will offer "somewhere between [a] 500% & 2000%" increase in its vehicle's autonomous driving performance. Existing Tesla owners who have already paid for full self-driving will be offered this "hardware 3" update for Autopilot free of charge. The announcement comes as v9 of Tesla's onboard software has already reportedly brought big improvements to its neural network with a unified camera network that more seamlessly integrates all eight of the car's cameras. Musk has suggested that this software update delivered an approximate 400 percent increase in its capabilities.

Google's tiny tensor processing unit (TPU) chips are shown perched on a pair of dice. Google, one of the top companies in the hot area of artificial intelligence, will begin letting customers directly use its custom processors for the technology starting in October. Google's TPUs, or tensor processing units, accelerate AI tasks like understanding voice commands or recognizing objects in photos. Today, Google will let you pay to do that kind of work on its cloud-computing infrastructure. But through a program called Edge TPU announced Wednesday, Google will let programmers install the TPUs in their own machines.

American depositary receipts of contract electronics manufacturer Taiwan Semiconductor Manufacturing (TSM) are down 29 cents, or 0.8%, to $36.27, despite an upbeat note today Susquehanna Financial Group's Mehdi Hosseini, who raises his rating on the stock to "Positive" from Neutral, after concluding the company should benefit in a big way from a wave of "custom" silicon being developed for artificial intelligence. That should help TSM "diversify" its customer base, which has traditionally been dominated by a few large customers such as Apple (AAPL) (TSM makes iPhone and iPad chips that Apple designs); Nvidia (NVDA); HiSilicon, the chip design arm of China's telecom giant Huawei; Qualcomm (QCOM); and Taiwanese wireless chip maker MediaTek (2454TW). TSM competes with Samsung Electronics (005930KS) for business from all those companies. "AI has had its fair share of hype for several years," he concedes. "But design activity from a diverse set of companies has finally started to increase materially: from VC-funded fabless players to the big CSPs (i.e., Google, Azure) with TSM winning most, if not all. "Thus, we are finally migrating from the era of smartphone/consumer to AI, which is mostly commercial app driven." AI is going to produce "incremental" revenue outside those big customers, writes Hosseini: Our recent checks, particularly as they relate to design and tape-out activities, suggest AI could account for as much as 20-25% of the overall revenues by 2020. The two groups of AI customers are: 1) leading edge, and 2) trailing edge [including] Cloud Service Providers, Private Cloud, OEMs, and Fabless design houses. The most striking takeaway is that only a few names within the two groups of categories are existing customers […] Looking forward, we expect incremental AI-related revenue contribution (materializing by mid-2019) to yield changes in TSM's top customer mix while helping with overall revenue diversification. Hosseini also thinks 5G wireless will be a boost to TSM: "5G remains a meaningful catalyst for higher mobile-related revenues by the 2020 timeframe.

Amazon was an early adopter of AI, and is working on custom chips that would have the ability to do on-device processing -- or at-the-edge processing -- rather than relying solely on a devices connection to the cloud, according to a report in The Information. Currently, when a user makes a request to Amazon's digital assistant Alexa, the information is transmitted to the cloud, which processes the request and submits a response back to the device. The ability to handle speech recognition locally would improve the response times for any device powered by the digital assistant, including the Echo family of smart speakers.

Amazon has started designing a custom artificial intelligence chip that would power future Echo devices and improve the quality and response time of its Alexa voice assistant, according to a report today from The Information. The move closely follows rivals Apple and Google, both of which have already developed and deployed custom AI hardware at various scales. AI tasks, because they are so computationally intensive, often need custom-designed chips for the devices themselves and even custom-designed servers for data centers where AI algorithms are often trained, developed, and deployed from the cloud. While Amazon is unlikely to physically produce the chips, given its lack of both fabrication experience and a manufacturing presence in China, the news does pose a risk to the businesses of companies like Nvidia and Intel. Both companies have shifted large portions of their chipmaking expertise to AI and the future of the burgeoning field, and both make money by designing and manufacturing chips for companies like Apple, Amazon, and others.

Today, Microsoft announced that the next generation of its mixed reality HoloLens headset will incorporate an AI chip. This custom silicon -- a "coprocessor" designed but not manufactured by Microsoft -- will be used to analyze visual data directly on the device, saving time by not uploading it to the cloud. The result, says Microsoft, will be quicker performance on the HoloLens 2, while keeping the device as mobile as possible. The announcement follows a trend among Silicon Valley's biggest tech companies, which are now scrambling to meet the computational demands of contemporary AI. Today's mobile devices, where AI is going to be used more frequently, simply aren't built to handle these sorts of programs, and when they're asked, the result is usually slower performance or a burned-out battery (or both).

Artificial intelligence has begun seeping its way into every tech product and service. Now, companies are changing the underlying hardware to accommodate this shift. Apple is the latest company creating a dedicated AI processing chip to speed up the AI algorithms and save battery life on its devices, according to Bloomberg. The Bloomberg report said the chip is internally known as the Apple Neural Engine and will be used to assist devices for facial and speech recognition tasks. The latest iPhone 7 runs some of its AI tasks (mostly related to photographer) using the image signal processor and the graphics processing unit integrated on its A10 Fusion chip.