With all the doom and gloom surrounding PC hardware in 2025, it's natural to wonder what comes next. Upcoming CPUs and GPUs might get all the attention, but I'm looking forward to emerging hardware standards in the PC and server space. These revolutionary trends promise to upend the underlying infrastructure powering CPUs, RAM, and SSDs, bringing unprecedented performance, efficiency, and scalability to consumer hardware. It might take a while to get there, but these rumblings are exciting enough.

NVMe-oF: NVMe-over-Fabrics

Going beyond PCIe

The current crop of NVMe SSDs that you're familiar with utilizes NVMe-over-PCIe technology to communicate with your computer. Well, enterprises are already using NVMe-over-Fabrics, or NVMe-oC, to swap the PCIe interconnect with Ethernet, fiber, TCP/IP, InfiniBand, and other alternatives. NVMe-oF is designed for use in server hardware, and its primary objective is to allow multiple devices to leverage a high-speed NVMe storage pool. Where NVMe-over-PCIe is limited to a single device, NVMe-oC effectively scales multiple SSDs to an entire network over long distances.

This new storage standard is not very relevant to consumer hardware, but it's fascinating, nonetheless, to see the flexibility at work. A single vast storage pool powering numerous hosts over non-PCIe interconnects feels like the natural evolution of the NVMe standard.

GAAFET: Gate-All-Around Field-Effect Transistor

Wrap it all around

GAAFET is the next step in transistor design, succeeding FinFET (Fin Field-Effect Transistor). As smaller nodes became prevalent in chip design, the challenges of FinFET gave way to GAAFET, which overcomes some of the limitations of the latter technology. Unlike FinFET, where the gate wraps around the channel on three sides, GAAFET has the gate surrounding the channel on all four sides. This 360º gate minimizes leakages, improves performance, and increases power efficiency.

When foundries started progressing beyond 3nm process nodes, it became necessary to transition to GAAFET. Samsung, TSMC, and Intel have their own versions of GAAFET for their 2nm nodes. Intel's RibbonFET technology is already operational in its 18A process powering the upcoming Panther Lake processors. These next-gen CPUs are poised to feature denser chips, higher performance per watt, and improved minimum voltage operation compared to Arrow Lake processors. AMD has confirmed plans to use Samsung's GAAFET technology for its upcoming products.

BPD: Backside Power Delivery

Making room for denser chips

Backside power delivery, in addition to GAAFET, is one of the pillars in modern chip design, allowing improved power integrity, lower power consumption, and better transistor utilization. Contrary to traditional processors, where both the signal and power wiring are located on the front side of the wafer, BPD relocates the power wiring to the backside. This frees up more space on the front side for denser signal wiring, improving yield and reducing the power delivery path. This separation also promises higher frequencies and reduced power loss.

Intel's Panther Lake CPUs make use of BPD in the form of its PowerVIA technology. The strength of its new 18A process node comes not only from PowerVIA, but also RibbonFET, a rebranding of GAAFET. Backside Power Delivery doesn't pose any cost or reliability challenges either, making it a bankable option for almost all upcoming chips.

CAMM2: Compression Attached Memory Module 2

Time to get low

CAMM2 is the next-gen form factor for memory modules, soon to replace the DIMMs and SO-DIMMs we're used to. JEDEC has already announced that DDR6 and LPDDR6 RAM will feature the CAMM2 form factor, relocating memory modules parallel to the motherboard. This radical change in memory form factor owes a lot to the signaling issues in DDR5 RAM when you cross certain speeds. Even empty DIMM slots can create signal interference that can hurt RAM performance.

CAMM2 solves this by eliminating soldered connections from the motherboard and moving the entire topology to the CAMM2 module. The benefits of moving from DIMM to CAMM2 are immense: a dual-channel 128GB configuration on a single module, better signal stability due to the direct connection to the motherboard, and enhanced compatibility with mobile devices due to the flat form factor. CAMM2 modules will seemingly make blazing-fast DDR6 speeds more stable, allowing the "sweet spot" for PC users to climb higher than ever before.

CXL: Compute Express Link

It could be here faster than you think

CXL, or Compute Express Link, is a fairly recent high-speed interconnect that links the CPU to memory and other accelerators to enable faster computing. It uses the physical PCIe connection, but builds on top of it with new protocols. CXL-compatible models of CPUs, GPUs, RAM, and NVMe SSDs can connect to a single large memory pool and use the CPU cache more efficiently. CXL is already being used in enterprise systems, thanks to CXL-compatible processors like AMD's EPYC Genoa and Intel's Sapphire Rapids.

While CXL is limited to server environments right now, it could conceivably make its way to consumer systems later. The benefits of an open protocol connecting all kinds of components and peripherals are immense. Companies like AMD even claimed that CXL support for consumer CPUs could arrive as early as 2027.

The future of PC hardware may not be so far away

Thanks to many of these standards already in operation in enterprise systems, their consumer hardware debut could be nearer than most people think. Intel's Panther Lake CPUs are already using GAAFET and BPD technologies, DDR6 has been confirmed to arrive on CAMM2 modules, and CXL support on CPUs could be here by the time PCIe 6.0 devices become available. It's an exciting time to be following these developments — who knows how things will evolve in the next 2–4 years.