Montreal, August 3, 2026 — Welcome back to Tech Science Daily, PcHybrid's plain-language tour of the engineering stories shaping the devices you buy. This week, three threads dominate the news cycle. First, the AI "memory supercycle" continues to redirect the world's supply of DRAM toward data centers, with real consequences for laptop, phone and SSD prices through the rest of 2026. Second, display science took a leap: LG Display's Tandem WOLED architecture is posting brightness and longevity numbers that would have sounded like fantasy three years ago, while RGB-LED backlights arrive to challenge OLED from the LCD side. Third, Microsoft's July Patch Tuesday shipped fixes for a staggering 570 vulnerabilities — including three zero-days and a SharePoint remote-code-execution flaw already being exploited in the wild. As always, we explain the underlying science, then translate it into practical buying advice with products you can actually get today.

Today's Tech Radar

The ten stories we shortlisted this week, before choosing our three deep dives:

Deep Dive 1 — The Great Memory Squeeze: Why AI Is Eating the World's RAM

A DRAM module up close: billions of one-transistor, one-capacitor cells. Photo: Liam Briese / Unsplash.

The science: what DRAM actually is

Every gigabyte of the memory in your laptop or phone is built from an astonishingly simple repeating unit: one transistor and one capacitor. The capacitor stores a bit as electric charge — charged means 1, empty means 0 — and the transistor acts as a gate that lets the memory controller read or change that charge. The catch is that capacitors leak. Within tens of milliseconds the charge fades, so the memory chip must constantly re-read and re-write every cell, thousands of times per second. That is the "dynamic" in Dynamic Random Access Memory. It is also why DRAM is cheap per bit compared with the static memory inside a CPU: a single transistor-plus-capacitor cell is tiny, so manufacturers can pack billions onto one die.

High Bandwidth Memory (HBM), the variant that AI accelerators crave, uses the same basic cell but changes the packaging entirely. Instead of spreading chips flat along a memory stick, HBM stacks up to a dozen DRAM dies vertically and connects them with thousands of microscopic vertical wires called through-silicon vias (TSVs). The result is a memory cube that sits millimetres from the GPU on a shared silicon interposer, with a bus thousands of bits wide. Bandwidth goes up by an order of magnitude; distance-related energy loss goes down. For AI models, whose main bottleneck is streaming billions of weights per second into the processor, HBM is not a luxury — it is the whole game.

Why your next laptop costs more

Here is the problem: HBM is made on the same production lines, from the same wafers, by the same three companies — Samsung, SK Hynix and Micron, which together control more than 95% of global DRAM output. HBM dies are bigger per bit, yields are lower, and the 3D stacking step adds cost and time. Industry analysts estimate that producing one bit of HBM consumes roughly the wafer capacity of several bits of ordinary laptop DRAM. So when the memory makers reallocate lines toward HBM — and with AI data centers projected by IDC and others to absorb around 70% of high-end DRAM output in 2026, they have every incentive to — the supply of conventional memory for laptops, desktops and phones contracts sharply.

The numbers this year have been dramatic. DRAM prices surged roughly 90% in the first quarter of 2026 compared with late 2025, and reporting from Tom's Hardware this week indicates contract prices for conventional DRAM are still expected to climb another 13–18% quarter-over-quarter in Q3, even as consumer demand cools at the edges. Retail RAM kits have jumped as much as 89% year-over-year by some trackers. NAND flash — the storage in SSDs and phones — is riding the same wave, because AI inference clusters also gobble enterprise SSDs. Gartner has warned of memory cost surges up to 130%, SK Hynix executives have suggested tightness could persist toward the end of the decade, and most analysts see elevated pricing through at least 2027 as new fabs take years to build.

Industry context: a supercycle, not a blip

Veterans of the PC industry have seen memory shortages before — they are cyclical by nature, because fabs take three years to build and demand moves in months. What makes 2026 different is that the marginal buyer is no longer a price-sensitive PC maker but a hyperscale AI operator for whom memory is a small line item next to GPUs and electricity. Alphabet alone has guided to roughly $195–205 billion in 2026 expenses, much of it data centers and custom silicon. When that kind of budget competes with consumer laptops for the same wafers, the laptops lose. IDC's analysis this summer projects the squeeze will visibly reshape smartphone and PC pricing and configurations into 2027 — expect base models with less RAM, fewer discounts, and longer replacement cycles.

What it means for buyers

The practical translation is simple: if you were planning a purchase that includes a meaningful amount of RAM or flash storage, earlier is very likely cheaper than later this year. A well-configured mid-range laptop bought today effectively locks in memory at prices that manufacturers negotiated months ago. At the store we currently have exactly one unit left of the Samsung Galaxy Book 4 15.6" (Core i5, 512 GB) (in stock) — a sensible, upgrade-friendly configuration whose 512 GB of NAND and full-size RAM would cost noticeably more to spec out at today's component prices. The same logic applies to storage upgrades: PCIe 4.0 drives like the Samsung 990 PRO 2 TB NVMe SSD (in stock) are exactly the class of component analysts expect to drift upward while the NAND market stays tight. And on the phone side, budget devices are where memory-cost pressure shows up first — the Samsung Galaxy A16 5G (128 GB) (in stock) remains one of the best value-per-gigabyte propositions on our shelves right now. If you are outfitting several workstations for a business and want help timing the purchase against the memory market, request a free quote from our team — we track component pricing weekly.

Deep Dive 2 — Stacked Light: How Tandem OLED and RGB-LED Backlights Are Rewriting Display Physics

Brightness, reflectance and lifespan — the three-way trade-off that tandem OLED architectures attack. Photo: BoliviaInteligente / Unsplash.

The science: why OLEDs age, and how stacking fixes it

An OLED pixel is a sandwich of organic semiconductor layers a few hundred nanometres thick. Drive a current through it and electrons meet "holes" (missing electrons) in the emission layer; each meeting releases a photon. The elegance is that every pixel makes its own light — no backlight, perfect blacks, instant response. The weakness is chemistry: the organic molecules degrade a little every time they emit, and the degradation accelerates disproportionately as you push more current. Roughly speaking, doubling the drive current far more than doubles the wear. That is why OLED TVs historically limited full-screen brightness — protecting the panel meant rationing the light.

Tandem architecture attacks this trade-off head-on. Instead of one emission layer, a tandem OLED vertically stacks two or more complete emission units, connected in series by a transparent "charge generation layer." Each stack now needs to supply only a fraction of the light, so each runs at a fraction of the current — and because degradation scales super-linearly with current, halving the per-layer load much more than doubles the lifespan. LG Display, which brands the approach Tandem WOLED, showed third-generation tandem panels at SID Display Week 2026 claiming 18% better power efficiency and roughly double the longevity of previous generations. Its 2026 flagship TV panels quote peak brightness of 4,500 nits — territory once reserved for premium mini-LED LCDs — combined with an industry-leading 0.3% reflectance, achieved with advanced anti-reflective films that suppress the mirror-like glare that used to plague glossy screens in bright rooms.

The counter-attack from LCD: RGB-LED backlights

Meanwhile, LCD is not standing still. The intriguing debut of 2026 — seen in models from Hisense, Samsung and LG Electronics, with Sony expected to follow — is the RGB-LED backlight. A conventional LCD backlight uses blue LEDs plus a phosphor or quantum-dot film to make white light, which the liquid-crystal layer and color filters then carve into an image. An RGB-LED backlight instead uses thousands of tiny red, green and blue LEDs whose color mix can be tuned zone by zone. Because the light arriving at the color filter is already close to the target color, less of it is thrown away, efficiency rises, and the achievable color volume — how saturated colors remain at high brightness — expands dramatically. Analysts still expect OLED and QD-OLED to lead the premium market in 2026 on contrast and viewing angles, with MicroLED remaining a niche halo technology due to manufacturing cost, but the LCD-versus-OLED race is now genuinely two-sided again.

Trickle-down: your phone got flagship-TV specs

A striking footnote from this month's phone launches: mid-range devices arriving in August quote 144 Hz AMOLED panels with peak brightness up to 5,000 nits (Vivo V80, iQOO Z11) and even a 7-inch 120 Hz OLED at 1,800 nits on a budget Poco. Panel science developed for televisions now reaches pocket devices within a couple of product cycles — the same tandem and efficiency tricks, miniaturized.

What it means for buyers

For most shoppers the lesson is that "4K panel" tells you almost nothing anymore — brightness, reflectance and duty rating are where the real differences live. If you want a large, bright 4K screen for a living space, lobby or meeting room, commercial-grade panels are an underrated option because they are engineered for long duty cycles at sustained brightness — exactly the failure mode that ages consumer screens. Our Samsung 55" Crystal UHD QBC (in stock) is a slim 4K display rated for 16/7 operation, and for truly big canvases the Samsung QM85C 85" UHD (in stock) delivers 500-nit non-glare output rated for 24/7 use — the anti-glare coating applies the same reflectance science discussed above. For desk work where color accuracy matters, the Samsung ViewFinity S8 32" 4K monitor (in stock) is our pick, and gamers chasing motion clarity should look at the Samsung Odyssey G5 34" 165 Hz curved monitor (in stock). Not sure which panel class fits your room and lighting? Request a free quote from our team and we'll match the spec sheet to your actual use case.

Deep Dive 3 — 570 Patches, Three Zero-Days: Anatomy of a Record Patch Tuesday

Patching is the cheapest security control there is — when it actually gets done. Photo: FlyD / Unsplash.

The news

Microsoft's July 2026 Patch Tuesday was one of the largest in the company's history: fixes for roughly 570 vulnerabilities, including three zero-days — flaws that attackers were exploiting before a patch existed. Separately, a SharePoint remote-code-execution vulnerability tracked as CVE-2026-45659 was added to CISA's Known Exploited Vulnerabilities (KEV) catalog after confirmed in-the-wild attacks, and a public exploit appeared for a Linux kernel flaw (CVE-2026-53264, CVSS 7.8) that lets any local user escalate to root. July's threat reporting also highlighted campaigns abusing legitimate Microsoft 365 workflows to bypass authentication controls, and a malvertising operation dubbed "SourTrade" that assembles its final malware payload inside the victim's own browser to evade download scanning.

The science: how a computer gets taken over remotely

"Remote code execution" sounds abstract, so here is the mechanism in plain terms. All software is a set of rules for transforming input into behavior. An RCE vulnerability is a spot where cleverly malformed input — a crafted web request, a poisoned document — pushes the program outside its rules so that part of the attacker's input gets treated as instructions rather than data. In server software like SharePoint, one classic route is unsafe deserialization: the server receives a package of data, reconstructs it into live program objects, and a crafted package smuggles in an object whose reconstruction runs attacker code. Because collaboration servers are, by design, reachable and trusted inside an organization, one such flaw can hand over the whole document store.

The Linux bug illustrates a second archetype: the use-after-free race condition. Modern kernels juggle thousands of simultaneous operations. If two threads touch the same piece of memory and one frees it while the other still holds a pointer to it, an attacker who wins the timing "race" can spray the freed spot with their own data — so when the stale pointer is used, the kernel executes attacker-controlled content with the highest privilege on the machine. Race conditions are notoriously hard to find in testing precisely because they depend on microsecond timing, which is why they keep surfacing even in thirty-year-old, heavily audited codebases.

Why patch volume matters — and why patching lags

570 fixes in a month is not a sign Microsoft's software suddenly got worse; it reflects an industry-wide surge in vulnerability discovery, increasingly accelerated by AI-assisted code analysis on both the defense and the attack side. The security economics are brutal and simple: the median time from patch release to weaponized exploit keeps shrinking, while the KEV catalog exists precisely because thousands of organizations run software with fixes available but not applied. For a small business, the attacker isn't a movie hacker targeting you personally — it is an automated scanner sweeping the internet for whoever patched last.

What it means for you

Three practical takeaways. First, turn on automatic updates everywhere — Windows, browsers, phones, even TVs; the July batch is exactly why. Second, retire hardware that no longer receives updates: an unsupported device is a permanent unpatched vulnerability, and with Windows 10's end of support behind us, machines that cannot run a current OS are structurally exposed. Modern hardware also brings NPU-accelerated, on-device security features — the new Panther Lake generation of AI PCs (story #8) runs threat detection locally at up to 180 platform TOPS. Third, businesses should treat patch management as a process, not an event: inventory, test, deploy, verify. If you run a small fleet and want an assessment of where your exposure is — aging machines, unsupported systems, or a patching workflow that needs automating — request a free quote from our team; a security-focused refresh often costs less than a single incident response engagement.

Glossary of the Week

Setup at a Glance

That's the radar for this week. If any of these stories raises a question about your own setup — from timing a memory-heavy purchase to hardening a small office against the current wave of exploits — request a free quote from our team and we'll build a recommendation around your actual needs.

Sources & Further Reading

Memory market: Tom's Hardware — memory prices climbing through Q3 2026; Wccftech — RAM shortage 2026 explained; IDC — global memory shortage crisis; Avnet — riding the AI supercycle; Tech Insider — SK Hynix warns shortage may last past 2030. Displays: LG Display — Tandem WOLED; Tom's Guide — next-gen Tandem WOLED; T3 — third-gen tandem OLED at SID 2026; FlatpanelsHD — RGB LED vs OLED; Analytics Insight — display tech in 2026. Security: BleepingComputer — July 2026 Patch Tuesday; Rapid7 — Patch Tuesday analysis; The Hacker News — SharePoint RCE added to CISA KEV; Xage — July 2026 risk roundup. AI PCs and chips: Engadget — Intel Panther Lake launch; Neowin — Core Ultra 300 details; TechCrunch — Google's new AI chip for Gemini. Phones: Gizmochina — August 2026 launches; Digit — smartphone launches to watch. Photos: Unsplash (free commercial license).