A new forecast from International Data Corporation (IDC) shows that enterprises will invest nearly $16 billion worldwide on GenAI solutions in 2023. This spending, which includes GenAI software as well as related infrastructure hardware and IT/business services, is expected to reach $143 billion in 2027 with a compound annual growth rate (CAGR) of 73.3% over the 2023-2027 forecast period. This is more than twice the rate of growth in overall AI spending and almost 13 times greater than the CAGR for worldwide IT spending over the same period.

"Generative AI is more than a fleeting trend or mere hype. It is a transformative technology with far-reaching implications and business impact," says Ritu Jyoti, group vice president, Worldwide Artificial Intelligence and Automation market research and advisory services at IDC. "With ethical and responsible implementation, GenAI is poised to reshape industries, changing the way we work, play, and interact with the world."

According to the editorial blog post published on the Samsung blog by SangJoon Hwang, Executive Vice President and Head of the DRAM Product & Technology Team at Samsung Electronics, we have information that High-Bandwidth Memory 4 (HBM4) is coming in 2025. In the recent timeline of HBM development, we saw the first appearance of HBM memory in 2015 with the AMD Radeon R9 Fury X. The second-generation HBM2 appeared with NVIDIA Tesla P100 in 2016, and the third-generation HBM3 saw the light of the day with NVIDIA Hopper GH100 GPU in 2022. Currently, Samsung has developed 9.8 Gbps HBM3E memory, which will start sampling to customers soon.

However, Samsung is more ambitious with development timelines this time, and the company expects to announce HBM4 in 2025, possibly with commercial products in the same calendar year. Interestingly, the HBM4 memory will have some technology optimized for high thermal properties, such as non-conductive film (NCF) assembly and hybrid copper bonding (HCB). The NCF is a polymer layer that enhances the stability of micro bumps and TSVs in the chip, so memory solder bump dies are protected from shock. Hybrid copper bonding is an advanced semiconductor packaging method that creates direct copper-to-copper connections between semiconductor components, enabling high-density, 3D-like packaging. It offers high I/O density, enhanced bandwidth, and improved power efficiency. It uses a copper layer as a conductor and oxide insulator instead of regular micro bumps to increase the connection density needed for HBM-like structures.

The latest Canalys data reveals that PC shipments (desktops, notebooks and workstations) in the US declined just 6% year-on-year to 18.2 million units in Q2 2023, marking a significant improvement compared to earlier quarters this year. Notebook (including mobile workstations) shipments were down 4% to 15.2 million units, bolstered by the return of Chromebook demand in the education sector. Desktops (including desktop workstations) suffered a steeper decline, with shipments falling 12% to 3.0 million units. The US tablet market faced a similarly modest decline, with shipments down 5% to 10.3 million units.

"Despite undergoing another year-on-year decline, the US PC market showed promising signs of improvement in the second quarter," said Ishan Dutt, Principal Analyst at Canalys. "With the buildup of channel inventories now largely cleared, pockets of demand strength are now being reflected in vendors' sell-in shipment performance. A key area that helped drive volumes was the return of demand from education institutions, backed by the latest wave of federal funding, ahead of a licensing cost increase for ChromeOS. This helped propel Chromebook shipments to 4.7 million units, the highest level since peak deployments during the first half of 2021."

According to TechNews.tw, TSMC could postpone its 2 nm semiconductor manufacturing node for 2026. If the rumors about TSMC's delayed 2 nm production schedule are accurate, the implications could reverberate throughout the semiconductor industry. TSMC's alleged hesitancy could be driven by multiple factors, including the architectural shift from FinFET to Gate-All-Around (GAA) and potential challenges related to scaling down to 2 nm. The company is a crucial player in this space, and a delay could offer opportunities for competitors like Samsung, which has already transitioned to GAA transistor architecture for its 3 nm chips. Given the massive demand for advanced nodes due to the rise of AI, IoT, and other next-gen technologies, it is surprising to hear "sluggish" demand reports.

However, it's also possible that it's too early for customers to make firm commitments for 2025 and beyond. TSMC has dismissed these rumors, stating that construction is progressing according to plan, which includes having 2 nm pilot run in 2024, and mass production in the second half of 2025.. Despite this, any delay in TSMC's roadmap could serve as a catalyst for shifts in market dynamics. Companies that rely heavily on TSMC's advanced nodes might need to reassess their timelines and strategies. Moreover, if Samsung can capitalize on this opportunity, it could somewhat level the playing field. As of now, though, it's essential to approach these rumors with caution until more concrete information becomes available.

Hardware leaker, YuuKi_AnS, has briefly turned their attention away from all things Team Blue—their latest leak points to upcoming server-grade processors chez AMD. A Zen 6 core-based 9006 EPYC CPU series, codenamed "Venice," is expected to arrive within two to three years along with an all-new SP7 socket—this information seems to have been sourced from an unnamed server manufacturer's product roadmap. A partial view of said slide also reveals forthcoming equipment powered by Intel "Falcon Shore" and NVIDIA "Blackwell" GPU technologies.

As reported a couple of months ago, older insider info has AMD using "Weisshorn" as an in-house moniker for Zen 6 "Morpheus" architecture, destined for Venice CPUs—alleged to form part of a 2025/2026 EPYC lineup. YuuKi_AnS proposes that these will utilize either 12-channel or 16-channel DDR5 memory configurations—thus providing plenty of bandwidth across hundreds of Zen cores. Altogether very handy for cloud, enterprise, and HPC workloads—industry experts reckon that 384-core counts are feasible on single packages. Naturally, a Team Red timeline dictates that Zen 5 "Nirvana" is due before Zen 6 "Morpheus," so EPYC 9005 "Turin(-X)" and 8005 "Turin-Dense" lineups are (allegedly) up for a 2024-ish launch window on SP5 (LGA-6096) and SP6 (LGA 4094) socket types.

Intel's integrated device manufacturing (IDM) has been experiencing a lot of trouble in recent years, and the company is not a leading-edge semiconductor manufacturer, with TSMC taking the pole position. However, the new restructuring hopes to change some of the business operations to increase its efficiency and establish Intel as the go-to foundry for customers. David Zinsner, Executive Vice President and the Chief Financial Officer, alongside Jason Grebe, Corporate Vice President & GM of the Corporate Planning Group at Intel, joined investors to explain how IDM will transform into a next-generation business. Intel IDM, including Intel Foundry Services (IFS), will get a new operation model, which will put IDM as an almost separate business unit with its own profit and loss (P&L) statement published in the quarterly/yearly financial report.

According to Intel, the company's IDM 1.0 strategy has been serving it well, but IDM 2.0 is needed to build next-generation nodes as the capital required for them is massive. Intel hopes to regain node leadership with the Intel 18A node in 2025. The company's strategy is still to have IFS as the second biggest external foundry business, presumably just behind TSMC. Putting IDM into its own P&L will result in $8-10 billion in "cost reduction opportunities, " including ramp rates, test time, and sort times based on the market pricing, not Intel's pricing. At the start, IDM is expected o start with a negative operating margin. Intel also states that keeping IFS as a business unit allows the company to simultaneously develop products on it and de-risk it for customers who want to build on IFS. The company is developing five different products (assuming packaging) on Intel 18A, all of which will be available for customers to use as well.

Chinese tech tipster Enthusiast Citizen (ECSM) has posted on Bilibili about future Intel product refreshes with a rough timeline spanning from late to 2023 to early 2024. We have been hearing a lot lately about Team Blue's Raptor Lake Refresh, with reports from this week suggesting that this lineup will be the last to sport Team Blue's traditional naming scheme—as 14th Gen Core. ECSM claims that Raptor Lake Refresh-K SKUs are due for launch this October, and non-K units will follow them a month or two later. Team Blue will likely be happy to keep LGA 1700 and 1800 sockets alive for another generation.

ECSM also brings up seemingly new information with an alleged Sapphire Rapids Refresh lined up for early 2024—suggesting that updated HEDT Xeon W2500 and W3500 series processors are incoming. The Intel W790 chipset should be able to run this rumored replacement lineup. ECSM's proposed product roadmap also presents wholly new product ranges including Meteor Lake offerings, albeit with desktop MTL-S SKUs cancelled—mobile-oriented Meteor Lake-H seems to be alive and well with an alleged Q4 2023 launch window. Finally Arrow Lake-S is predicted to launch in the final quarter of 2024 or early 2025—so we will likely have to wait another year and a half for upgraded Intel mainstream desktop SKUs.

During the ISC High Performance 2023 international conference, Intel announced interesting roadmap updates to its high-performance computing (HPC) and artificial intelligence (AI). With the scrapping of Rialto Bridge and Lancaster Sound, Intel merged these accelerator lines into Falcon Shores processor for HPC and AI, initially claiming to be a CPU+GPU solution on a single package. However, during the ISC 2023 talk, the company forced a change of plans, and now, Falcon Shores is GPU only solution destined for a 2025 launch. Originally, Intel wanted to combine x86-64 cores with Xe GPU to form an "XPU" module that powers HPC and AI workloads. However, Intel did not see a point in forcing customers to choose between specific CPU-to-GPU core ratios that would need to be in an XPU accelerator. Instead, a regular GPU solution paired with a separate CPU is the choice of Intel for now. In the future, as workloads get more defined, XPU solutions are still a possibility, just delayed from what was originally intended.

Regarding Intel's Gaudi accelerators, the story is about to end. The company originally paid two billion US Dollars for Habana Labs and its Gaudi hardware. However, Intel now plans to stop the Gaudi development as a standalone accelerator and instead use the IP to integrate it into its Falcon Shores GPU. Using modular, tile-based architecture, the Falcon Shores GPU features standard ethernet switching, up to 288 GB of HBM3 running at 9.8 TB/s throughput, I/O optimized for scaling, and support for FP8 and FP16 floating point precision needed for AI and other workloads. As noted, the creation of XPU was premature, and now, the initial Falcon Shores GPU will become an accelerator for HPC, AI, and a mix of both, depending on a specific application. You can see the roadmap below for more information.

TrendForce forecasts a global downturn in AR and VR device shipments for 2023, predicting a shipment total of roughly 7.45 million units—an 18.2% YoY decrease. VR devices are expected to shoulder the majority of this decline, with projected shipments hovering around 6.67 million units. This dip is primarily attributed to weaker-than-expected sales of newly released high-end devices. Consequently, manufacturers are likely to pivot their sales strategies, shifting their focus to more cost-effective offerings. Conversely, shipments of AR devices are expected to remain stable, with projected shipments exceeding 780,000 units. While Apple's latest offerings could stimulate some demand, the high price tags attached to these units continue to pose a significant barrier to broader market growth.

Two key factors emerge when examining the impending decline in shipments in the VR sector. Firstly, brands may have been overly optimistic regarding the sales of their top-tier products. Despite these premium devices offering enhanced features courtesy of advanced hardware and software, consumers are showing reluctance to shoulder the associated higher costs. Instead, they seem to be gravitating towards more budget-friendly models this year. Secondly, the shortage of appealing, new, cost-effective models in the market is exacerbating the downturn. Meta Quest 2 continues to maintain its status as this year's market-leading VR product as the release of Meta Quest 3 has been pushed back to 2024.

A leaked roadmap (that seems to be authored) by Giga Computing provides an interesting peak into the future of next generation enterprise-oriented CPUs and GPUs. TDP details of Intel, AMD and NVIDIA hardware are featured within the presentation slide - and all indications point to a trend of continued power consumption growth. Intel's server CPU lineups, including fourth generation Sapphire Rapids-SP and fifth-gen Emerald Rapids-SP Xeon chips, are projected to hit maximum TGPs of 350 W by mid-2024. Team Blue's sixth gen Granite Rapids is expected to arrive in the latter half of 2024, and Gigabyte's leaked roadmap points to a push into 500 W territories going forward into 2025.

AMD's Zen 5-based Turin server CPUs are expected to ship by the second half of 2024, and power consumption is estimated to hit a maximum of 600 W - representing a 50% increase over the Zen 4-based Genoa family. The 2024 NVIDIA PCIe GPU lineup is likely hitting TDPs of up to 500 W, it is rumored that these enterprise cards will be based on the Blackwell chip architecture - set to succeed current generation H100 "Hopper" PCIe accelerators (featuring 350-450 W TDPs). It is possible that AMD's Instinct-class PCIe accelerator family will become the direct competition, these cards are rated up to 400 W. The AMD Instinct MI250 OAM category has a maximum rating of 560 W. The NVIDIA Grace and Grace Hopper CPU Superchips are said to feature 600 W and 1000 W TDPs (respectively).

In a rather unexpected move, Samsung will reportedly open a chip development facility in Yokohama, Japan. According to the Nikkei, Samsung is readying a 30 billion yen or US$222 million investment near its current R&D institute in Japan. Samsung is hoping to be able to leverage a combination of Japanese and Korean expertise at the site, although exactly what kind of chip development that will take place at the site is currently unknown, beyond it being focused on the back-end processor of chip manufacturing. This generally involves the wafer packaging process or chip stacking, processes that have evolved a lot of the past few years.

The facility is said to be employing hundreds of people once it starts operating sometime in 2025. The Nikkei is also reporting that Samsung is hoping to take advantage of subsidies offered by the Japanese government, which might also be one of the reasons for opening the development unit in Japan. The subsidies are said to be in excess of 10 billion yen. Considering that Japan and Korea aren't on the best terms at the moment, for many reasons and most of them irrelevant to this news post, it's surprising to see Samsung making this move, as although it might be a fairly minor investment for the company, it's doing so on what could only be referred to as hostile soil.

The US Government is considering new plans that will attempt to curb the after effects of cryptocurrency mining. The White House revealed details about its proposed "DAME Tax" scheme on Tuesday of this week - the Digital Asset Mining Energy excise tax is under consideration for this year's US Budget. The government wants to address the impact that cryptomining has on the US economy as well as the environment, alongside numerous other national challenges. Companies engaged in the extraction of cryptocurrencies could be charged extra for the running of computer equipment (starting in early 2024). A White House spokesperson states: "after a phase-in period, firms would face a tax equal to 30 percent of the cost of the electricity they use in cryptomining."

American crypto companies are facing a 10 percent taxation of their energy bill for 2024, that will then increase to 20 percent in 2025, and the maximum tax rate will hit a high of 30 percent in 2026. The White House number crunching team reckons that $3.5 billion could be generated by the proposed DAME excise tax. The new rules would represent a radical change for large scale cryptomining efforts: "Currently, cryptomining firms do not have to pay for the full cost they impose on others, in the form of local environmental pollution, higher energy prices, and the impacts of increased greenhouse gas emissions on the climate. The DAME tax encourages firms to start taking better account of the harms they impose on society," reads a White House statement." The government's investigation has determined that the domestic cryptomining industry is close to consuming more electricity than the entire nation's residential lighting system. US lawmakers last year calculated that some of the larger digital asset mining firms are capable of using more energy than nearly all of the residential population based in Houston, TX.

Microsoft has confirmed that the current version of Windows 10 - 22H2 - will be the final one. A company product manager revealed this information yesterday in a Windows IT Pro Blog entry posted alongside a mass of articles on Microsoft's Tech Community site. As covered on TPU almost two years ago, Microsoft had given advance notice that it was terminating support for Windows 10 on October 14th 2025 - for both Home and Pro versions of the operating system. Windows 11 was released later on in 2021, and thus became the priority OS product for the North American tech firm.

Yesterday's blog reiterates key information from the past, and details an interim update cycle (albeit small): "Windows 10 will reach end of support on October 14, 2025. The current version, 22H2, will be the final version of Windows 10, and all editions will remain in support with monthly security update releases through that date. Existing (enterprise) releases will continue to receive updates beyond that date based on their specific lifecycles." Microsoft has proceeded to update the lifecycle page entry for Windows 10 Home and Pro in line with the latest announcement. The product manager (in his blog) recommends that current Windows 10 users move to 11 as soon as possible, in order to enjoy a continued stream of feature updates.

AMD and JEDEC are collaborating to create a new industry standard for DDR5 memory called MRDIMMs (multi-ranked buffered DIMMs). The constant need for bandwidth in server systems provides trouble that can not easily be solved. Adding more memory is difficult, as motherboards can only get so big. Incorporating on-package memory solutions like HBM is expensive and can only scale to a specific memory capacity. However, engineers of JEDEC, with the help of AMD, have come to make a new standard that will try and solve this challenge using the new MRDIMM technology. The concept of MRDIMM is, on paper, straightforward. It combines two DDR5 DIMMs on a single module to effectively double the bandwidth. Specifically, if you take two DDR5 DIMMs running at 4,400 MT/s and connect them to create a single DIMM, you get 8,800 MT/s speeds on a single module. To efficiently use it, a special data mux or buffer will effectively take two Double Data Rate (DDR) DIMMs and convert them into Quad Data Rate (QDR) DIMMs.

The design also allows simultaneous access to both ranks of memory, thanks to the added mux. First-generation MRDIMMs can produce speeds of up to 8,800 MT/s, while the second and third generations modules can go to 12,800 MT/s and 17,600 MT/s, respectively. We expect third-generation MRDIMMs after 2030, so the project is still far away. Additionally, Intel has a similar solution called Multiplexer Combined Ranks DIMM (MCRDIMM) which uses a similar approach. However, Intel's technology is expected to see the light of the day as early as 2024/2025 and beyond the generation of servers, with Granite Rapids likely representing a contender for this technology. SK Hynix already makes MCRDIMMs, and you can see the demonstration of the approach below.

All eyes - especially investors' eyes - are on Intel's data center business today. Intel's Sandra Rivera, Greg Lavender and Lisa Spelman hosted a webinar focused on the company's Data Center and Artificial Intelligence business unit. They offered a big update on Intel's latest market forecasts, hardware plans and the way Intel is empowering developers with software.

Executives dished out updates on Intel's data center business for investors. This included disclosures about future generations of Intel Xeon chips, progress updates on 4th Gen Intel Xeon Scalable processors (code-named Sapphire Rapids) and demos of Intel hardware tackling the competition, heavy AI workloads and more.

Xeon Roadmap Roll Call
Among Sapphire Rapids, Emerald Rapids, Sierra Forest and Granite Rapids, there is a lot going on in the server CPU business. Here's your Xeon roadmap updates in order of appearance: