CNBC reports Barclays raised its price target for Intel shares in a note to clients this week. The company expressed its view that Intel will emerge as a leader in artificial intelligence and other next-generation technologies. Although Intel is generally perceived to be playing catch-up in AI, Barclays is confident that the chipmaker can shape the future of the concept.
Return of the CPUs
The predictions centre around the hardware Intel uses to power AI applications. Much of the industry has moved towards using GPUs to train and operate neural networks. GPUs offer considerable performance advantages over regular processors (CPUs) when running machine learning workloads. The trend towards purchasing GPUs, predominantly supplied by NVIDIA, has come at the expense of CPU makers such as Intel.
In its research note, Barclays predicted that CPUs will gain increased relevance again over the next few years. This is because the improved performance of GPUs doesn’t come without its trade-offs. In many real-world scenarios, CPUs still offer better operating speeds and a more controllable environment.
They’re also much easier to program and use, offering developers a more familiar experience with a shallower learning curve than GPU-based AI. With AI experts still in short supply, the accessibility advantages of CPUs might see them installed in more applications as their technology progresses.
Performance improvements for new apps
Intel hasn’t been sitting still with its CPU technology. Recent platform releases have delivered dramatic improvements to computational performance which have directly impacted machine learning workflows.
2015’s Skylake architecture doubled AI performance in some benchmarks and has been followed-up by further gains with Kaby Lake and Coffee Lake. Although innovation in CPUs is becoming increasingly difficult, Intel’s still finding ways to optimise its chips to run neural networks.
READ NEXT: Microsoft uses “neural fuzzing” technique to find software bugs
Earlier this month, Intel announced a new industry-first silicon chip for neural network processing. It’s part of the company’s aim to create AI “100 times faster” than last year’s tech before 2020.
This will help Intel consolidate its position in the market and power “new classes” of AI application. The company’s particularly targeting AI-powered vehicles, healthcare and customer experience services. Intel said it’s on course to reach $46 billion in AI revenue by 2020.
CNBC reports Barclays sees Intel’s stock rallying by another 20 percent as its tech finds its way into more AI services. Its share price is up 26 percent this year. The analysts said the effect of Intel’s revitalised AI portfolio should become apparent starting in 2018.
