EDACafe Industry Predictions for 2024 – Alphawave Semi

Tony Pialis

Vertical customization will save AI hardware’s limits.

• With the launch of ChatGPT in November 2022, 2023 was the “year of AI,” spurring companies of all sizes and industries to dive headfirst into the AI race. With that came an exponential need for hardware to support this data-intensive technology more efficiently and cost-effectively. Microsoft, AWS, and even the developer of ChatGPT, OpenAI, are actively developing or have announced they are looking into developing their own custom silicon solutions to support their AI applications. In 2024, we will continue to see hyperscalers and enterprise companies who are working with AI at scale develop their own AI-focused custom silicon chips to support their AI applications.
  • The only effective way for companies to bring their custom designs to market fast and without creating a globally splintered ecosystem is through the use of chiplet-based designs, utilizing standards such as Universal Chiplet Interconnect Express (UCIe), as companies search for the most advanced and compatible chiplet designs for their existing hardware products. We will also see a rise in inference-focused chips to support the growing need for rapid inference as more users worldwide start to utilize AI applications.

Hardware will become the hottest, most in-demand tech sector (going head-to-head with AI). 

• Software has long been the “sexiest” tech – heavy emphasis has been placed on coding boot camps, it’s been perceived as a greater opportunity for wealth creation, and more interest and investment dollars have gone into this versus its hardware counterpart.
• However, as a more volatile economy has seen many SaaS companies downsize, the introduction of the CHIPS Act, and the unyielding demand for AI chips means that semiconductors – and hardware – is once at the top of the list of the most in-demand – and “recession proof” tech.
• 2023 saw the newest addition to the “trillion dollar club” as NVIDIA was the seventh company to reach a market cap of $1 trillion in June and easily outperformed all other stocks on the S&P 500, as it worked to meet the insatiable demand for AI-application supporting hardware. Even towards the end of the year, the company has maintained its $1T market cap, serving as a sign for semiconductor companies that increased investment in these critical technologies is here to stay as we continue to meet the needs of AI applications in the future. As more semiconductor companies, like Arm, get closer to joining its rank.
• While much has been written about AI’s explosion in popularity, no AI systems can run without sophisticated data infrastructure. AI is the gold rush, but the folks who make hardware are the ones selling the shovels.
  

AI will face a large-scale reckoning–and the cream will rise to the top

• AI interest has exploded in 2023, with companies quickly embracing/adopting various applications, and ~10 million Americans already start their regular web searches using AI tools.
• However, the rapid and widespread adoption of artificial intelligence is already hitting some snags – e.g. ChatGPT has seen declining user numbers, and complaints are rife that its inference models are becoming less accurate.
• The challenges AI is seeing in accuracy, neutrality, and ethical use are continuing to bubble up each day – and as we continue to use AI, it’ll be imperative to address and resolve these problems. Otherwise, these will snowball in volume and velocity of issues as AI continues to scale.
• In 2024, hardware/software vendors & government organizations will need to continue to work more closely together to prioritize the right set of investments, safeguards, and innovation to build the right foundation for AI and minimize any business/tech challenges as AI inevitably scales.
• In the past year alone, thousands of AI projects were funded. There just isn’t a market to support all the companies getting funding, and the data infrastructure is not where it needs to be to optimize their technologies–but that’s a good thing. Like the dot com bubble of the early ‘00’s, only the companies with truly sterling products with invaluable use cases will survive and–as our data processing infrastructure catches up–become part of our day-to-day

2024 will be the year that kicks off the optical revolution–and pushes forth a brave new world

• For years, promises of new-age Jetson-esque tech advancements have been hindered, in large part, due to limitations within our data infrastructure. Your phone could download an entire HD movie in two seconds, if it wanted to, but the data infra on the backend just can’t support that–but optical can. While this has been known for years, the large-scale explosion of AI has made it so that 2024 will be the year large industry players invest at large in optical, kicking off a decade in which tech advancements will seemingly come a mile a minute.
• Experiments in optical networking have produced incredible results in recent years. Still, the next generation of optical networking will be unleashed in 2024 thanks to the unstoppable push to increase connectivity speeds thanks to the explosion of AI.
• We will see:
  • The development of solutions that blow past 200Gbps/wavelength to increase the data rates to keep up with the insatiable connectivity needs of AI.
  • Key players start to push for adopting standards for AI-specific connectivity solutions to ensure interoperability for future generations of optical networking technologies.
  • We will see the beginning of an all-optical connectivity world thanks to recent advancements in co-packaged optics and similar technologies to increase data rates into the future.

About Author:

This entry was posted on Monday, January 1st, 2024 at 5:41 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.