主管觀點部落格

2022 年和之後的預期趨勢:人工智慧、產業 4.0 和永續發展科技將要求新的運算典範

作者 David Moore - 2022-01-31

By David Moore, Chief Strategy Officer and Senior Vice President of the Corporate Strategy and Communications organization at Micron

At the start of a new year, it’s customary for Micron to take stock of what the next year (and the years after) will hold. While no one truly has a crystal ball into the future, we are continuously looking at how technology usage is transforming industries and how the role of memory and storage is evolving to become more central to that transformation. Understanding the new ways the world will generate, consume, interact with and manage growing amounts of data helps us to stay ahead of the curve and drive not only business results, but progress for the planet and its people.  

As access to data becomes even more pervasive – enabling emerging usage models and incredibly rich, interactive experiences – we’ll see greater demands for the technologies we use daily and the cloud to edge infrastructure needed to deliver it.  

These demands, in turn, will require unprecedented levels of technology innovation for a growing array of devices that we use to interact with our data at home and on the move. These technologies will need to:  

  • Deliver exceptional performance and incredible data capacity
  • Enable new and breakthrough levels of energy efficiency at the system level 
  • Harness the opportunity that artificial intelligence (AI) provides to transform user experience and business results
  • Drive progress in the industry against critical reusability and sustainability objectives  

In 2021, we announced significant investments in manufacturing, R&D, sustainability and many more initiatives to help meet the continued growth in demand for innovative memory and storage technology. In 2022, we must continue to drive innovation in the face of that exponential demand across industries for high-bandwidth, energy-efficient solutions — from devices to the intelligent edge to the cloud.

In my role as chief strategy officer, I get to work across our many teams as we chart and hone our strategy for Micron. In the process, I am fortunate to engage with outstanding business thought-leaders and visionary technologists from across Micron. We’ve asked them to share their perspectives on key market dynamics we expect to emerge and predict what technology trends will take hold in 2022 and beyond.

The Data Center of the Future

Raj Hazra, senior vice president and general manager of Micron’s Compute & Networking Business Unit

The data center – on-premises or cloud – underpins our digital lives. It’s the infrastructure that hosts all the services we know and love; everything from online shopping and inventory tracking to real-time map navigation and content streaming. As the world’s insatiable demand for data and new and improved services grows, data centers are increasingly pushed to their limits, unable to store and process all available data efficiently. 2022 will see the beginning of immense transitions for data center architectures, with memory and storage driving much of it. This is the era of composable architecture, with service providers and enterprises increasingly focused on sustainability and with data security an ever-present concern.

A new memory and storage hierarchy is bringing a revolution to data centers in 2022. With data center infrastructure under stress, innovation across various types of memory will be key to unleashing data-intensive workloads. The most exciting transition is the ramp of DDR5. This is a generational leap in performance and the first time data center needs are truly driving memory standards. We’ll also see the emergence of the first Compute Express Link (CXL) based memory solutions, an initial step on the path to data center composability and memory pooling. It’s also the year we will welcome HBM3, which when paired with general purpose central processing units (CPUs), gives a huge boost for in-package memory. For the data center, these innovations will result in the emergence of a new memory hierarchy that we haven’t had before and a more composable infrastructure with pools of compute, memory and storage. For enterprises and service providers alike, that may mean better performance and lower total cost of ownership across many types of workloads and applications.

The move toward less rigid, more composable infrastructure will also enable better energy efficiency. By reducing overprovisioned resources, data center operators can reduce their environmental footprint. According to the U.S. Department of Energy, the largest data centers require over 100 megawatts of power – enough to power about 80,000 households. With 8.5 million data centers on earth, these facilities have a drastic impact on energy and water consumption, carbon emissions and waste. Expect companies to get more strategic about sustainability. They will closely look at how they manage their infrastructure in energy-efficient ways as well as take a stronger position on green technologies that make up that infrastructure, such as low-power memory and storage. Companies looking at data center sustainability through a strategic business lens can no longer afford to approach memory and storage as just components. Rather, they will look at them as key enablers of their business and sustainability goals and a true composable infrastructure.

Security will be about more than data encryption. Data encryption is widely recognized and increasingly implemented, but security is also about reliability and availability of data and services. It is not just about encrypting or protecting bits and selected devices; it is about having key data and services secure and always available. A true data center security plan now must take into account securing the data itself not only from breaches, ransomware, etc., but also phishing attacks and social engineering, as well as disaster recovery. In addition, we’ll see the rise in automation and self-healing to alleviate the effects of data center breaches. Expect service providers and enterprises to continue to take this wider view of security, allowing their customers and employees to readily intermingle their physical and virtual lives while keeping data centers running optimally and securely.

An artist painting on a large canvas

Augmented Reality, Virtual Reality, and Now onto “Surreality”

Dinesh Bahal, corporate vice president and general manager of Micron’s Commercial and Components Group 

Metaverse or not, augmented reality (AR) and virtual reality (VR) will drive a convergence of client devices. There is significant discussion in the market about how deeply and how completely people engage in metaverse-like virtual worlds. Which is why the term “surreality” suggested by one of the leading industry analysts is so apt for the era we’re entering. There is no debate that increasingly pervasive and immersive digital experiences will become some part of many people's daily lives, and the devices that enable them will continue to evolve to meet user experience expectations. Look for even more vertical integration across your devices, such as more VR built into smartphones and more AR built into smart glasses. And many of the devices we use daily will have increased integration of technologies such as haptics to engage more of our senses and create multi-sensory, physical experiences. Within these systems, expect more innovation in chip packaging and other items traditionally outside the memory or storage die to help enable more flexible form factors.

For the last 40 years, people have been looking at things in a two-dimensional plane, whether it's a computer or a phone screen. VR has been a slow technology to scale up and out, but now everybody's talking about a multi-dimensional immersive experience. Until now, resolution, responsiveness and latency have been bottlenecks to mixed reality going mainstream, but these capabilities are finally at a point where VR and immersive experiences are becoming viable. And what that will do to compute is exciting from a memory and storage perspective.

Mixed reality will gain traction, where we’ll see the merging of real and virtual worlds to produce new environments and visualizations. These new experiences will accelerate with more untethered VR systems allowing users to move around unencumbered and provide much more immersive experiences in the VR world. The realization of 5G at scale and the emergence of the intelligent edge will open up this opportunity like never before. You'll also have more and more content being pushed into untethered devices, so you’ll need even higher performance, more graphics rendering, and lower power memory and storage in a compact space. You'll see unique form factors in memory, like what you see in the mobile space – and they’ll continue to need more storage capability, more bandwidth and more memory.

Sustainability and right to repair will take flight. The world produces 20 to 50 million metric tons of electronic waste each year with only a fraction of that properly recycled. But now some of the leading smartphone companies are allowing consumers to fix their own phones without breaking warranty. This is going to give rise to a movement across all technology manufacturers where sustainability, reusability and upgradeability are going to become bigger pieces of the story – don't buy a new PC or device, upgrade your existing products. That will accelerate an already burgeoning notion of the circular economy, which instead of looking at a product as at its end of life, aims to eliminate waste by reusing, refurbishing, repairing and recycling existing materials. In 2022 and beyond, we’ll see other manufacturers of devices following suit and engaging in a more sustainable future.

The arm of a robot

A Transformation of the Supply Chain and Automotive Architectures at the Intelligent Edge

Kris Baxter, corporate vice president and general manager of Micron’s Embedded Business Unit

With the proliferation of data and insights from billions of devices comes the rise of the intelligent edge. The industry is recognizing the importance of processing data where it is created for better insight – not all use cases can tolerate the latency or cost of going back to a cloud data center for processing. The intelligent edge is increasingly seen as a viable strategic investment. Enterprises are seeing the benefit of real time-analytics for predictive maintenance and factory optimization as well as creating more efficient cities, making efficient use of digitalization.

The transformation of the intelligent edge is being accelerated with the advancements of AI and 5G, and accordingly, the benefits of connectivity and compute are extending into environments like industrial applications and advanced automotive experiences. As edge deployments mature, they bring new requirements for high-performing memory and storage in challenging environments such as extended life spans, expanded operational temperatures and space and energy optimizations.

Industry 4.0 will accelerate. In 2022, amidst continuing supply constraints, many organizations are going to confirm that making their infrastructure and applications more intelligent will help maximize their output from their supply chain network. Digital twins will help them optimize manufacturing and minimize supply chain shortfalls, and capabilities like computer vision and acoustic sensors will quickly detect manufacturing flaws or issues on the manufacturing floor. The need for optimization of manufacturing will accelerate Industry 4.0 — an opportunity that has been studied for a while, but that many enterprises are just beginning to fully realize. Continued advancements in connectivity, WiFi6 and 5G, along with advanced AI chips and software, will drive a growth in smart gateways, 5G private networks, edge data centers and smarter devices with enhanced management and orchestration capabilities.

We’ll see faster automotive technology migration. Increasing adoption of advanced driver assistance features, consumers’ expectations for up-to-date user experiences combined with a strained global supply chain are collectively pushing the automotive industry into a new mindset. In the coming year we will see accelerating migrations from legacy technology that is supply-starved to next-generation platforms that have new and innovative products and solutions available. A decade ago, the adoption and validation cycle of technology from mainstream non-auto applications into automotive-grade products was about five years, now that’s down to about one and a half years. This can only aid the acceleration of next-gen tech in the automotive industry.

We’ll also see the automotive industry increasingly embrace centralized vehicle architectures. The U.S. has been lagging in the electric vehicle (EV) adoption behind China and Europe, but in the next year, we will see an acceleration of EV rollout and U.S. adoption. This creates an exciting opportunity for vehicle designers to accelerate the introduction of new architectures such as centralized or zonal architecture, where centralized memory, storage and compute versus a highly distributed architecture will drive greater efficiencies. These architectures will change requirements for memory and storage to higher capacities with higher-bandwidth capabilities in order to support the increasing need for data creation, storage and processing within the vehicle.

Automotive and industrial will be two of the fastest growing segments in the coming years, perfectly illustrating how connectivity and intelligence are enabling new use cases in which data is processed where it is created for better user experiences.

Mobile devices recording a live presentation

 

Your Phone Will Understand You Better

Raj Talluri, senior vice president and general manager of Micron’s Mobile Business Unit

Mobile phones have become powerful personalized devices, providing tailored services and information through various apps and features. This ability to understand the user, situation or different environmental variables, called contextual awareness, will continue to get more robust as more sensing and compute options become available.

Today, people don’t even think about the fact that their phone has multiple sensors, cameras and other context-aware hardware already built in. While the most commonly known sensor is GPS, there are several other sensors on your smartphone such as a gyroscope, a pedometer, a magnetometer and an altimeter, to name a few. This hardware is critical for creating a unique experience for the user. In 2022, we’ll see the device and app ecosystem take further advantage of these capabilities to create unique user experiences.

5G and AI will bring contextual awareness to life. This year, we expect to see an acceleration of contextual awareness and human-centered AI as interconnected objects will be infused with AI and empowered to better sense the environment and your patterns with its many sensors. Imagine, a morning commute where your alarm goes off and your coffee machine knows exactly how long you take to get out of bed to make a cup of coffee. Then your phone tells you when it’s time to leave for work based on your calendar entries, and informs your car to power on, pull up to your driveway, and take you to the office with the most efficient route by communicating with other cars, city cameras and traffic lights. While in the car, your watch monitors your emotions and mood and plays music custom tailored for you until you arrive. This is the future with 5G and AI central to its success.

Your smartphone will become more intuitive. The current imaging technology built into smartphones has made capturing quality images seamless and automatic. When a user points the camera and zooms in on a subject, the phone automatically switches from one lens to another. In low-light situations that require longer shutter speeds, your phone’s sensor compensates for your shaky hand and helps capture a sharp image. Also, many phones are employing 3D sensing or lidar for depth management, so they can blur parts of the image for selective focus. While most of these technologies are available in flagship phones today, in 2022 and beyond, these capabilities will make their way further down the stack and be available in all phones at all tiers for all users. And it’s just a matter of time before most of the back of the phone will be all cameras!

Memory and storage growth will outpace other components. All of this mobile technology will drive a dramatic increase in the amount of memory and storage required in mobile phones. As we expected, the amount of memory and storage in phones has essentially doubled with the transition from 4G to 5G. The average memory density smartphones shipped with was approximately 5.2 GB in 2021. We believe this will grow to an average of 9 GB of memory required per device in 2025.

Sound waves

AI: Continuing the Chase for Brain-Level Efficiency

Steve Pawlowski, Micron corporate vice president of advanced computing solutions & emerging memory solutions

The primary approach to AI will increasingly shift (back) to AI-enhanced hardware—and that will not just impact processors. It is comparable to the birth of the Internet and all those railroad crazes of the past – in the beginning, there was euphoria and a number of small companies circling to build the winning solution. Ultimately, the market sorts out where real value and thus the best solutions lie. Similarly, when AI started entering mainstream consciousness, we saw the same thing: everybody had their new AI or neural processing chip and companies were proclaiming that AI processing was going to be central to everything we do. Of course, most of these radical approaches failed to find a market, and the general trend has been to throw loads of general purpose or graphics processing unit-centric compute at the AI opportunity. That works for a while, but scales poorly and is not cost-effective.

Now as knowledge grows across industries, there’s a more pragmatic understanding in the market that Rome wasn’t built in a day. For AI to take the next step to becoming ubiquitous, it needs to be efficiently processable which means that there needs to be more of a focus on AI-enhanced hardware — and that does not mean going back to the discrete AI or neural chip approach. Rather, in 2022 and beyond, expect AI solutions to be more balanced and blended – just as we see AI techniques being added to or paired with virtually any kind of application, from computer-aided design to customer service to retail and communications.

For 5G networks, most service providers expect that they can gain efficiency by using traffic data and AI algorithms to train the networks to do a better job of beam steering with multiple antennas. Increased efficiency in 5G implementations would be significant when one considers the cost of 5G baseband and antenna infrastructure and spectrum! Much of this 5G work is happening on open platforms and through open standards. It is an improvement, but not perfect, there are many experiences with occasional bottlenecks and resource contentions as these new open systems balance AI and network processing roles. Micron is currently working to analyze the challenges in this space and develop opportunities to improve the efficiency of AI and digital baseband algorithms to help the industry accelerate the transition to open 5G.

The industry is already looking at the next “G”– 6G. With this we’ll see an increased focus on memory-centric, AI-based, digital algorithms, which will drive the growth of the memory and storage footprint for these new wireless technologies. We’re part of the industry collaboration scoping requirements and capabilities for 6G networks now, so we can make sure that memory and storage technologies are ready to deliver the throughput, capacity and low latency these networks, devices and user experiences will demand. But we’ll temper 6G predictions for now and focus on 5G!

As AI solutions evolve, we’ll take our cues from the amazing computing device that is the human brain. Electric circuits still use five to six orders of magnitude more energy than biological systems. It’s amazing the amount of work that a human brain does on roughly 35 watts!

Today’s big complex systems such as ADAS are completely insignificant compared to what the brain can do. The brain is basically a large memory device that does a remarkable job of computation, and a big reason for that is fixed function areas. The area of the brain that handles vision is much different than the area that handles hearing or other cognitive tasks. A brain is basically a large heterogeneous system where there are different fixed function applications with high performance connectivity to bring it all together.

The same concept is needed for AI – specialized sub-systems for specific tasks, as opposed to more massive general compute. This will allow better optimization of energy usage, which is desperately needed as AI continues to use astronomical amounts of energy. For example, OpenAI found that training a language model used almost the same amount of energy three homes use in a year! Clearly, if AI is going to be everywhere, it needs to be much more efficient.

A technician working in a server room

The Next Victim of the Cloud: Hard Disk Drives

Jeremy Werner, corporate vice president and general manager of Micron’s Storage Business Unit

Very high capacity and low-cost solid-state drives (SSDs) will accelerate the extinction of hard drives. The performance growth profile of SSDs has outpaced the performance curve of CPUs. Of course, this leads people to ask: how do you get the most out of this flash performance? The first stop for most is to share that performance across more systems to improve performance utilization. But as more systems share drives, the capacity requirements of the drives increase, leading to demand for larger capacity drives. As high-performance NVMe-over-fabric shared storage becomes increasingly prevalent in cloud and composable infrastructures, we’ll see the demand for high-capacity and low-cost SSDs increasing – replacing shared disk drives in many cases due to the overall total cost of ownership savings of deploying flash versus disk.

These savings in terms of power, space, secondary hardware consolidation, reliability and relative lifespan are well accepted today, but with increased high-performance storage disaggregation, the trend will accelerate in 2022. Hard disk drives on the other hand are struggling to meet baseline levels of performance today and don’t scale performance well with capacity as SSDs do, so over the next decade SSDs will essentially relegate hard drives to the same archival-like, low-performance use cases many use tape for today.

PCIe Gen5 will have its biggest and most immediate impact on network connectivity and accelerators. The industry typically awaits generational change technologies such as PCIe Gen5 with great expectations – and perhaps a bit of hype. But we all know that architectural transitions take time, and often don’t deliver value or get deployed immediately or linearly. This will be true for PCIe Gen5. We expect that the most pressing bottlenecks in today’s data centers are frequently related to processing power limits – particularly for very parallel and specialized workloads such as machine learning and AI model training that drive a need for application-optimized accelerator offload engines, and moving large amounts of data and files through the network, driving the adoption of faster networking equipment such as 100-, 200-, 400- or more gigabit networking. Higher performance PCIe connectivity is critical to solving those compute and networking bottlenecks that are prevalent and currently pressing issues.

We see less near-term urgency from enterprises for PCIe Gen5 as a critical storage enhancement because of the large amount of parallelism in the storage subsystems and the still-very-recent migration to PCIe Gen4. Again, this is logically related to the more pressing bottlenecks being in acceleration and input/output constraints. While PCIe Gen5 storage would indeed allow faster and more scalability in storage (always a good thing!), it generally doesn’t make sense to add cost to scale storage performance if the system can’t take advantage of the additional storage performance due to compute or I/O bottlenecks. Quite simply, if one put 20 or 24 high-performance Gen5 SSDs into a server, the performance impact at the solution level is low, because all that data can't get out to the network in the case of disaggregated storage or you can't process it fast enough with most CPUs in the case of direct attached storage. But once high-performance PCIe Gen5 based networking and acceleration is commonplace, organizations will more broadly be able to take advantage of the performance a PCIe Gen5 NVMe SSD can deliver with a solid return on investment. So maybe it's still early for PCIe Gen5 SSDs, but PCIe Gen5 network interface controllers and accelerators will unlock the value and stranded performance in PCIe Gen4 SSDs.

The pandemic will continue to firmly establish hybrid cloud and co-location as the most critical parts of a company’s IT strategy. No longer are people thinking of their IT as a binary choice between on-premises or cloud. Most companies have come to recognize that to get the performance, flexibility and control they need, a hybrid strategy is the way to go. Not too long ago, the prevailing thought was that public cloud would eventually consume all application and infrastructure needs and that private IT departments and data centers would become less relevant. And while the public cloud continues to grow and offer tremendous flexibility and capabilities to their customers, companies continue to heavily invest in data centers and in-house IT. It is clear that public cloud and private cloud are going to coexist moving forward.

Companies’ choices in cloud infrastructure are also evolving. Service providers are looking at ways to exist at the intelligent edge to deliver services that require lower latency and even data sovereignty. The challenge is that in the face of the pandemic, maintaining staff and developing the infrastructure to operate data centers (especially at the edge) is difficult. For this reason, expect that managed services data centers will continue to play a growing role in companies’ hybrid cloud strategies. Enterprises will grow to appreciate the balance they enable: offering material control over infrastructure – similar to on-premises, privately owned data centers – but without the need to have the staff or the know-how to maintain the hardware and build the data center.

The predictions made within this article are for informational and entertainment purposes only. Though these predictions are made by Micron experts, they do not constitute market guidance for the company or the markets of our customers.

David Moore

David Moore

David Moore is Chief Strategy Officer and Senior Vice President of the Corporate Strategy and Communications organization at Micron. In this capacity, he is responsible for leading Micron’s corporate strategy, M&A strategy, venture investments, and global communications and marketing.

Previous to Micron, David held various positions at Intel where he was most recently the corporate vice president and general manager of the Programmable Solutions Group, a $2 billion business unit. He began his career in engineering at Altera, purchased by Intel in 2015. During his almost 21-year tenure at Altera/Intel, he was based in Malaysia, Vietnam and the U.S., providing him a strong global outlook.

David earned a bachelor’s degree in electrical and electronics engineering at the University College Dublin and an MSEE in electrical, electronics and communications from the Georgia Institute of Technology.

+
Powered by Translations.com GlobalLink OneLink SoftwarePowered By OneLink