“Huixin Cup” Emerging technology + Internet Innovation Competition, sponsored by the Industrial Culture Development Center of the Ministry of Industry and Information Technology, the competition brings together senior venture capital institutions, more than 2 million cash awards and more than 100 well-known media, accelerating the landing of entrepreneurs’ incubation projects; No charge for entrywww.chuangcheng.org.cn/455

“Huixin Cup” Contest Consultation :13828821502(same micro phone number)

For years, financiers who invested in technology had little interest in startups making computer chips — it was hard to imagine how a start-up could compete with a giant like Intel. After all, more than 80% of the world’s personal computers run on chips made by Intel. Even in areas where Intel doesn’t dominate, such as smartphones and gaming devices, there are companies like Qualcomm and Nvidia that can crush the start-ups.

But then came the latest fad in the tech industry: artificial intelligence. It turns out that new computer chips can run artificial intelligence better. All of a sudden, venture capitalists seem to have forgotten the factors that keep young chip companies from succeeding.

Today, at least 45 start-ups are developing chips that can power tasks like speech recognition and self-driving cars, and at least five of them have raised more than $100 million from investors. Venture capitalists invested more than $1.5 billion in chip start-ups last year, nearly double the amount two years earlier, according to CB Insights, a research firm.

The surge is similar to the sudden expansion of PC and hard drive makers in the 1980s. While these companies are small, and not all of them will survive, they are capable of driving a period of rapid technological change.

It remains to be seen whether any companies fancy challenging Intel with their chips — Intel has spent billions of dollars building its own chip factories. (These startups contract with other companies to make chips.) But when designing chips to provide machine learning with the specific computing power it needs to perform an increasing number of tasks, the start-ups are moving in one of two directions: to quickly find a profitable niche, or to secure a big corporate acquisition.

“Machine learning and artificial intelligence are redefining the question of how to build computers,”

In the summer of 2016, changes were already evident. Google, Microsoft and other Internet giants are developing apps that use algorithms, known as neural networks, to recognize faces in photos and recognize commands on smartphones. These algorithms can learn tasks by identifying patterns in large amounts of data.

China has also shown particular interest in developing new AI chips. A third Beijing-based chip startup, DeePhi, has raised $40 million, and China’s Ministry of Science and Technology has explicitly ordered the production of Chinese chips to challenge Nvidia.

Because this is a new market — and because there is such a strong appetite for this new type of processing power — many see this as a rare opportunity for start-ups to go up against entrenched giants.

The first big change is likely to come in the data center, where companies like Graphcore and Cerebras have been quiet about their plans to accelerate the creation of new forms of ARTIFICIAL intelligence, including the ability to hold conversations, automatically generate video, and virtual reality.

Researchers at companies such as Microsoft and Google have built chips specifically for AI and repeatedly “train” neural networks through extreme experiments, testing algorithms on a large number of chips. The training process can last for hours or even days. They often sit in front of their laptops, staring at charts and watching these algorithms progress as they learn data. Chip designers hope to simplify the process, packaging all the trial and error into a few minutes.

In addition to building chips specifically for neural networks, start-ups are rethinking the hardware around them. Graphcore, for example, is building chips with more built-in memory to save the hassle of sending large amounts of data back and forth. Other companies are looking at ways to expand transmission between chips to exchange data more quickly.

“It’s not just about building the chips, it’s about studying how the chips are connected together and how they interact with the rest of the system,” sequoia’s Coughran said.

But that’s only part of the change. Once the neural network has been trained into a task, additional tools must be available to perform the task. At Toyota, prototypes of self-driving cars are using neural networks to identify pedestrians, signs and other objects on the road. After training a neural network in a data center, the company runs the algorithm on a chip installed in a car.

A number of chipmakers — including start-ups like Mythic, DeePhi and Horizon Robotics — are also tackling the problem, pushing AI chips into devices as diverse as phones and cars.

It is unclear how well the new chips will work. It takes about 24 months to design and manufacture chips, meaning that even the first viable hardware that relies on them won’t be available until this year. And those chip start-ups will face competition from Nvidia, Intel, Google and other industry giants.

But the starting point is the same: the beginning of a new market.