Amazon’s $1,349 15-inch MacBook Air deal delivers the lowest price of the season. Upgraded models are $150 off as well.

Save $150 on Apple’s 2026 15-inch MacBook Air, bringing the price of the M5 model with 16GB of RAM and 512GB of storage down to $1,349.

Buy M5 15″ MacBook Air for $1,349

The limited-time deal is available in the Midnight finish only, and units are shipping within 1-2 days, suggesting stock may be limited.

Amazon’s $150 discount extends to several upgraded configurations as well, with the M5/16GB/1TB model also marked down to $1,649 in the Midnight finish.

Our M5 MacBook Air Price Guide offers easy price comparison across popular retailers on dozens of configurations. Here are today’s highlights from the guide:

13-inch MacBook Air deals

• 13″ M5, 8C GPU, 16GB, 512GB: $1,239 ($60 off)
• 13″ M5, 10C GPU, 16GB, 1TB: $1,449 ($150 off)
• 13″ M5, 10C GPU, 24GB, 1TB: $1,699 ($100 off)

15-inch MacBook Air discounts

• 15″ M5, 10C GPU, 16GB, 512GB: $1,349 ($150 off)
• 15″ M5, 10C GPU, 16GB, 1TB: $1,649 ($150 off)
• 15″ M5, 10C GPU, 24GB, 1TB: $1,899 ($100 off)

You can also browse the best MacBook Air deals throughout the week in our dedicated roundup.

Memory prices are finally slowing down, and the reason is not that supply improved. It is that buyers ran out of money.

TrendForce expects conventional DRAM contract prices to rise 13% to 18% quarter on quarter in Q3. Dan Robinson reported the forecast for The Register. He paired it with separate European shipment numbers that show what the squeeze has done to the market.

The Taiwanese analyst calls that growth rate a moderation, and by recent standards it is. Rising contract prices drove a 59.5% quarter-on-quarter jump in DRAM industry revenue in Q2, to $154.73bn.

Why it is slowing

TrendForce gives two reasons, and the second is the interesting one.

Demand is shifting away from high-capacity RDIMMs towards lower-capacity parts. And then there is what the analyst calls “the limited ability of PC and smartphone customers to absorb further price increases”.

That is a market hitting a ceiling rather than finding a balance. Supplier inventories are still at historic lows and bit shipments are expected to grow only modestly. Nothing about the supply side has eased. Prices are rising more slowly because the people at the end of the chain cannot pay any more.

Note who TrendForce names as the constraint: PC and smartphone customers. Phones use the same memory and face the same maths, and the handset market has spent a decade training buyers to expect more storage each year for roughly the same money. That expectation is now colliding with the price of the parts.

And 13% to 18% in a single quarter is not relief. It is a smaller increase on top of a year of increases.

Europe is where the bill lands

The market intelligence firm Context expects European PC shipments to fall hard as component costs push system prices up.

Laptop shipments are forecast to drop 6.4% year on year in Q3, then 20% in Q4. Desktops look worse: down around 20% in Q3 and almost 30% in Q4.

Those are severe numbers for a mature market, and they describe a straightforward mechanism. Memory got more expensive, so machines got more expensive, so European buyers stopped replacing them.

Corporate buyers are stretching refresh cycles and buying only when they have no choice, according to Context. “The PC refresh cycle has not disappeared, but the economics around it have changed,” said senior analyst Marie-Christine Pygott.

Stretching a refresh cycle is not free, though it looks free on a budget line. Older machines run slower, fail more often, and eventually fall off support, which is a security problem rather than a productivity one. European IT departments are taking that risk because the alternative costs more today.

Falling volumes, rising profits

PC makers are not suffering, which is the part that explains why nothing corrects.

Higher prices have more than covered the lost volume so far. The Register reported in May that Lenovo had dodged the worst of the memory crisis by pushing customers towards premium devices. The strategy worked.

That is a rational response to scarcity and it also removes the pressure to fix it. If selling fewer, dearer machines makes more money than selling more cheap ones, a manufacturer has no reason to fight for volume.

The same logic has been visible up and down the market for months. AMD raised Radeon prices in August, days after Nvidia did the same. Nvidia is putting up AI server prices by more than 15% from next year, citing memory costs. Everyone passes it on.

The AI PC did not save anyone

Manufacturers spent two years promising that AI-capable PCs would trigger a mass upgrade. Context says that is not what happened.

The end of Windows 10 support drove recent buying, not artificial intelligence. AI PCs are taking a growing share of machines entering the European supply chain. That is mostly because the capability is becoming standard, though, rather than because anyone is choosing it.

So the upgrade wave came from a support deadline, and it has now largely passed. What is left is a market with higher prices, no compelling reason to buy, and a stack of machines that will stay in service longer.

Who the memory is going to instead

None of this is happening to the AI industry. It is happening because of it.

The three big memory makers are concentrating on high-end products for AI buyers. Samsung has benefited from moving early into HBM4 mass production. SK hynix had the highest proportion of HBM in its bit shipments of the three. Micron is prioritising higher-priced server DRAM.

That leaves the mature-process business to second-tier manufacturers such as Nanya, Winbond and PSMC, who are supplying DDR4 and DDR3. The squeeze has already reached parts nobody expected, including DDR2, a standard from 2003, and the capacitors that carmakers need.

Europe watches most of this from the outside. SK hynix chose Indiana for its $4bn memory packaging hub after a Polish plant fell through, which means the continent absorbs the price rises without hosting much of the capacity.

What would actually change it

Two things, and neither is close.

The first is supply. Inventories are at historic lows and bit shipments are growing modestly, so relief needs new capacity, which takes years to build and is currently being pointed at HBM rather than at the parts in a laptop.

The second is AI demand cooling, which nobody in the industry expects and which would bring its own problems.

Until one of those happens, the pattern holds. AI buyers pay a premium for the memory they want. Everyone else pays a premium for the memory that is left. And the moderation TrendForce is forecasting is a description of European buyers reaching their limit, not of the shortage ending.

The fuel for a fully reusable rocket that can unlock truly low-cost access to space isn’t methane or liquid oxygen. It’s cash, and Stoke Space Technologies is tanking up. The company has now “completed the initial closing” of a $1 billion Series E round intended to help it reach orbit and prepare a new, larger rocket for operations.

“This round is really to scale,” CEO Andy Lapsa told TechCrunch. “To lay the infrastructure, to scale in production and flight frequency, and importantly to fund the development of the second generation vehicle.”

The round was led by Point72 Ventures (hedge fund billionaire Steve Cohen’s tech investment vehicle), and Spark Capital, and included investment from General Innovation, Glade Brook Capital, US Innovation Technology, Washington Harbour Partners, Woven Capital and Y Combinator, among others.

Stoke is one of a handful of startups attempting to compete with Elon Musk’s SpaceX in the low-cost rocket launch space. SpaceX was built around the Falcon 9, which has a reusable booster stage, and now Musk and company are trying to field a huge, fully-reusable rocket, Starship.

But unlike its rivals Rocket Lab, Relativity Space or Firefly Aerospace, Lapsa’s company is aiming straight at that holy grail: A rocket where both the booster stage and the payload-carrying second stage come back to Earth so they can be reused cheaply. It’s never been done before.

Such ambition and technical feats require capital. SpaceX has spent well over $10 billion on Starship over the last decade, and the vehicle still hasn’t yet reached orbit, although that may change on its next flight. Now, Stoke has raised $2.3 billion in total.

In particular, SpaceX has struggled to identify the right materials and configuration for the components that shield the second stage of the rocket from the extreme heat caused when it is reentering the atmosphere. Stoke instead flows super-cooled liquid hydrogen through its thermal shielding, a unique active cooling solution that Lapsa says has been tested exhaustively on the ground and will be deployed on the first flight.

“We can flow excess coolant — more than we think we need in order to overcool the surface of the vehicle — and take a conservative stance from that perspective in early flights,” he said.

Lapsa said that he expects Stoke’s first vehicle, the Nova Pathfinder, to take flight in early 2027. In a business known for delays, he’s confident sharing that target following a series of structural and operational tests on the rocket’s first stage at the company’s Moses Lake facility. Next up is test-firing each stage of the rocket on the ground to ensure they are ready for flight.

“The ground system is checked out to the extent that we can do it without the rocket, and the rocket’s checked out to the extent that we can do it without the launch pad,” Lapsa said. “So the next step is to do them together.”

The startup has already sold launch contracts on the vehicle, which can carry three metric tons to low-Earth orbit — that will be, Lapsa believes, the largest debut vehicle of any U.S. rocket maker. That said, few rockets get off the ground on time or with total success, and simply getting to orbit would be an achievement.

“We have multiple Pathfinder vehicles in production, and we have confidence that we will bring Pathfinder to market and into orbit regardless of this round,” Lapsa said.

Stoke also revealed plans for an even bigger rocket, the Nova Block 2, that has been under development for the past few years. Based on the same technologies as the Pathfinder, including the engines and the company’s novel heat shield, that vehicle will be able to bring 15 metric tons to low-Earth orbit, slightly more than the Falcon 9.

The goal is to deploy it in 2029, right when, according to Musk, the Falcon 9 will be phased out. Is Lapsa drooling at the possibility of picking up increasingly desperate launch customers just as SpaceX pulls back?

“It amplifies the mismatch between launch supply and demand for launch, there’s no question about that,” he admits. “Regardless of Falcon 9 retiring or not retiring, the space industry and the space economy scales exactly as fast as rockets get off the ground, particularly rockets that serve third-party customers.”

That last remark is a dog whistle that satellite operators will surely hear: When SpaceX’s Starship takes flight, companies and government agencies looking to buy tickets will be competing with the space company’s internal projects, and might have to pay a premium to use the big rocket. Stoke, on the other hand, has not revealed plans to operate its own space assets.

“Now is the time to get moving […] so that we can finally start deploying a lot of the constellations and applications that we aspire to as an industry but have been stuck on the ground,” Lapsa said.

Appeals Centre Europe began providing dispute resolution services in Ireland in 2024.

UK-based independent press regulator Impress is launching a new service in Ireland to help take down illegal content, such as AI-generated deepfakes, from social media.

Impress Dispute Resolutions Services – Impress’ Dublin subsidiary – is an out-of-court dispute settlement (ODS) body certified by Ireland’s media regulator Coimisiún na Meán.

Launching today (8 September), it is offering a free of charge, independent mechanism for EU citizens to contest harmful content and platform decisions under the EU’s Digital Services Act (DSA). IDR’s scheme is available to users in Ireland and the EU and covers major platforms including Instagram, Facebook, TikTok and YouTube.

The DSA mandates that online platforms must provide a clear and accessible means to lodge complaints against illegal content. They must also allow users to request a review if they disagree with a platform’s decision.

In addition, the law also provides users the ability to turn to ODS bodies, which offer a faster and more cost-effective way to settle disputes with social media platforms when compared to the much more tedious judicial routes. These bodies are generally provide services at a low or no cost.

While ODS bodies cannot impose a binding settlement, platforms and users are required to engage with them in good faith.

“Over the past decade, we’ve supported thousands of people harmed in the media – and demand has never been higher,” said Impress Dispute Resolutions co-founder Gia Thom. “We’re non-profit by design, which keeps IDR independent, credible and focused on strengthening the information ecosystem.”

The EU has 11 listed ODS services spread across the bloc, including the Appeals Centre Europe, which began providing ODS services in Ireland in 2024.

Safety on social media has come to a head in recent years, with parents and lawmakers moving to shield children from the provable harms posed online by banning popular platforms.

Meanwhile, the proliferation of illegal content is getting compounded by AI, which enables bad actors to generate realistic-looking (often non-consensual and sexual) images and videos in seconds.

IDR, in addition to dealing with AI deepfakes, can also help users navigate disputes around illegal or harmful speech, the negative impacts on civic discourse or elections, and data protection, among other areas.

“Impress has delivered independent and effective redress for the public harmed by the media for over a decade, and we are very pleased to extend our proven service to those harmed by decisions of tech platforms,” said Impress CEO Lexie Kirkconnell Kawana.

“We know intimately and rise to the challenge of balancing the rights of individuals, groups and organisations while upholding the law and protecting free speech. Impress represents integrity in all that we do, and we will continue to embody this principle in providing fair outcomes for all participants to our scheme.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

A rumored “Silo” spinoff could keep Apple TV’s dystopian franchise alive beyond its planned finale, but the books’ author is pushing back on the claim.

“Silo” is a dystopian science-fiction drama about a society living underground after humanity has been driven from the surface. The Apple TV series is based on the bestselling “Silo” trilogy, “Wool,” “Shift,” and “Dust,” by Hugh Howey.

While it was well liked, Apple TV had told the creators that the series would end after the fourth season. The third season just returned to the small screen in July, with the fourth season anticipated to release in Summer 2027.

Even though the series is facing down its inevitable end, it may not be time to say goodbye just yet.

Allegedly, there’s a chance “Silo” may be getting its own spinoff, set to begin pre-production in late 2026. It is expected to tie into Hugh Howey’s upcoming “Silo 40” trilogy, which is set to release in 2027.

This tantalizing tidbit was shared by Apple TV enthusiast Sigmund Judge on Threads.

One minute later, however, a very significant figure refuted the claim. Speaking up was the “Silo” author, Hugh Howey.

“Yo, this is news to me,” Howey says in the replies. “So probably not true.”

“I appreciate you wanting to keep a surprise,” Judge says in response. “Love everything you do.”

Judge says that he’s got some further evidence in response to “a denial,” presumably Howey’s. Judge says that hires for the upcoming project have begun, and that the main sets used for “Silo” have remained, even after production on the fourth season wrapped in March.

HoverAir Aqua

MSRP $1,299.00

Advertisement

“A brilliant specialist with a short attention span.”

Pros

• Rights itself and launches from open water every single time
• Tracking at 15 mph is steady enough to pass for a chase boat
• The Lighthouse wearable handles launch, landing, and following
• Onboard screen lets you replay clips on the water with no phone
• Dead battery? It lands softly on the surface and floats

Cons

• I measured 10 and 13 minutes of flight against an advertised 23
• The Lighthouse shows no battery level at all
• No obstacle avoidance
• The battery is only fully waterproof while it sits in the drone

Quick Take

I spent two weeks on an Oregon river trying to break this thing — off an eFoil, out of a kayak, and by throwing it into the water on purpose. The HoverAir Aqua is the first consumer drone actually built to live on open water, and after two weeks I still stop what I’m doing to watch it launch. Throw it in the river upside down, and it flips upright, shakes the water off its props, pops into a low hover, and waits for you. Strap the Lighthouse beacon to your forearm, and it will follow you across the river at eFoil speed with your hands never touching a controller.

Let me be blunt, because this is the most important thing to know before you buy: there are two battery problems. HoverAir prints 23 minutes on the box. I measured 10 minutes and 13 minutes with the drone actively tracking me. And the wearable never tells you how much charge is left. So this is a drone I love, packed in a dry bag next to a fistful of spare batteries I resent having to carry.

HoverAir Aqua specs: A quick peek at the innards

HoverAir Aqua design & build: A pool toy with flagship chops

The first thing I noticed pulling the Aqua out of the box is the fat ring of safety-orange foam around its body, the same shade as a life vest. It looks like a pool toy. That’s deliberate. The foam keeps the drone buoyant, cushions the props, and makes the aircraft easy to spot from fifty yards out when it sets down on dark green water. The whole package carries an IP67 rating and weighs under 249 g, which keeps recreational pilots in the U.S. under the FAA registration threshold.

My favorite piece of hardware here is also the smallest. A 1.6-inch AMOLED screen sits on the drone’s spine, and it stays readable even with glare coming off the water. It’s there to pick flight modes, but I mostly used it to review what I’d just shot — sitting on the board, dripping, thumbing back through a run while the river went past. Two weeks in, I was still reaching for it instead of my phone. Storage is 128GB and internal, so there’s no SD card to lose, forget, or flood.

The camera records 4K at up to 100 fps for slow motion through a heated lens, which keeps fog from forming and makes water bead off instead of smearing across the glass. Underneath the airframe is the strangest line I have ever typed about a consumer drone: millimeter wave radar that reads the height of the waves rolling beneath it, so the Aqua can skim low over moving water without clipping a crest.

For those of you unfamiliar with the term, and since the spec sheets never bother explaining it, millimeter wave radar bounces very short-wavelength radio off a surface to measure distance to it — here, a surface that will not hold still. The industry noticed. The Aqua premiered at CES 2026 as an Innovation Awards honoree and collected a Red Dot Award along the way.

HoverAir Aqua water launch: Throw it in, it flies out

The water launch is the feature every bystander will ask you to do twice, and I tested it from the kayak with my kids leaning over the side to watch. You toss the drone into the river upside down, carelessly. However, it happens to leave your hand. It bobs for a beat. Then the props bite, it rolls upright, and it climbs into a hover a few feet off the surface to wait for you to start moving. It did that every single time I asked, and I asked a lot. Not once did I have to paddle over and fish it out or reset anything. In a category where water has always meant total loss, that is a bigger deal than a party trick sounds.

HoverAir Aqua subject tracking: A chase boat on your forearm

Most of my testing happened on my eFoil at around 15 miles per hour, with the waterproof Lighthouse strapped to my forearm like a chunky watch. One button sends the drone up. One button brings it home. Everything between those two presses is automatic.

The footage surprised me. My wake unspooling behind the board in a long white seam, the tree line sliding by, and me held in the center of the frame for the entire run. The best material came when the drone swung out ahead of the board and shot back at me, spray kicking off the foil, the whole river opening up behind.

It never fell behind, and it never wandered. When I drifted too far, it closed the gap on its own. When I slipped off the board and sat chest-deep in the water beside it, it stayed locked on and waited. The kayak sessions at lazier speeds went the same way: steady framing, no fuss, the family paddling along with a small orange aircraft holding formation off the bow. This is the closest thing to a camera operator you can wear on your wrist.

HoverAir Aqua Battery Life: High on the box, dips in the water

Now, the part the box will not tell you. HoverAir advertises 23 minutes of flight from the 2,013mAh smart battery. My first battery, actively tracking me on the foil, gave me 10 minutes before the drone called it quits. I rode back to shore, clicked in a fresh one, and went back out. Thirteen minutes. Doing the exact job you bought it for, you get about half the number on the sticker.

It’s worth noting that drone makers rate flight time in windless conditions with no tracking load, and you should not expect that figure in the real world. Even allowing for that, half is a wide gap. At least the ending is graceful.

When the charge runs out, you get a warning that the drone will land in 15 seconds, and then it sets down on the surface, orange side up, rocking on the chop until you come get it. Nothing dramatic happened in two weeks of testing. Recharging takes about 55 minutes with the battery inside the drone, and because a battery is only fully waterproof while it’s installed, HoverAir includes a dedicated waterproof pouch for hauling spares out on the water. Read that accessory list again. The company is telling you, in its own way, that you will be swapping batteries mid-session.

HoverAir Aqua Lighthouse Wearable: Great controller, no fuel gauge

The second problem makes the first one worse. The Lighthouse is otherwise the best thing about this product, and it tells you nothing about the battery. No percentage. No halfway warning. No blinking light as the level sinks. Mid-run, I had no idea whether the drone above me had eight minutes left or thirty seconds.

When the entire budget is 10 to 13 minutes, that silence turns every session into a guessing game. Both times a battery died on me, my only notice was the 15-second landing warning, right when I felt like I was getting going.

Should you buy

The good here is genuinely good. The footage is beautiful, the Aqua keeps pace with a foil at speed, it launches itself out of open water, and it runs hands-off enough that you forget it exists until you sit down to replay the clips. As a piece of engineering built for a hostile environment, it is the real deal.

The battery is what makes me hesitate, and I am not going to soften that. Plan a full session around this drone, and you are packing a pocketful of spares and guessing when to burn them.

So here is the fork. If what you want is a quick promo shot or a fast social clip off the water, this is the drone doing that job, and I don’t know of another one built to take the punishment. If you want to shoot an entire session, wait and watch for a firmware update that pushes battery data to the Lighthouse.

That one change would make this an easy recommendation for anyone who makes content on the water. As it stands, the Aqua is a brilliant specialist with a 10-minute attention span, and I keep taking it out anyway.

Why not try?

HOVERAir X1 PROMAX — Priced significantly lower at $700, this one offers hands-free auto-follow tracking and Lighthouse wearable perks, can shoot up to 4K 120fps footage, and stands out with a foldable design. It’s slightly slower and doesn’t offer a waterproof build, but makes up for it with collision-sensing tech.

DJI Mini 4 Pro — Another option that is more affordable as well, DJI’s drone offers impressive aerial capture capabilities with omnidirectional obstacle sensing and a reliable 4K camera that can capture high dynamic range videos. Plus, the battery packs can extend the flight time by up to 45 minutes, while the transmission performance is also fantastic. But an impending ban casts a shadow of skepticism over long-term support and after-sales experience.

FAQs (Frequently Asked Questions)

Who needs the HoverAir Aqua?

It’s a waterproof, self-flying 4K camera drone. If you’re into activities such as kayaking and wakeboarding, it’s an appropriate choice.

Is it capable of taking off and landing directly on water?

Yes, the positive buoyancy engineering allows it to float on water. And yes, it can take off and land from the surface of water. Moreover, it also comes with a Turtle Flip feature that lets it correct its position instantly, even if it’s flipped upside.

Is the build really waterproof?

The HoverAir Aqua has an IP67-certified build, featuring corrosion-resistant components, while the camera sensor is protected by a hydrophobic, anti-fog lens.

Do you need a license to fly it?

It falls under the prescribed FAA weight limit, so you don’t need a license to fly it.

Can it fly and capture without a phone?

Yes, the wearable Lighthouse controller relies on Real-Time Kinematics (RTK) positioning to let the drone track you and capture videos without requiring a cellular link or base stations.

What is the top speed and battery life?

The HoverAir Aqua can fly at a speed of up to 34 miles per hour, and it is claimed to last over 20 minutes on a single charge.

Can you fly it manually?

Yes, using the Beacon & JoySticks accessory.

In a nutshell: Despite Nvidia losing fans over the last few years as its focus on gamers diminishes (or disappears), the company’s GPUs have dominated the sales charts. But that status quo is starting to change: Amazon’s top 10 best-selling GPU list is an even split between Team Red and Team Green, with the highest position held by a Radeon 9000-series card. Meanwhile, in Germany, AMD is the one that’s dominating sales – at least at one retailer.

The rise of RDNA 4 was highlighted by TechEpiphany on X.

– TechEpiphany (@TechEpiphanyYT) September 6, 2026

The Amazon best-selling list has changed slightly since the post, but the Radeon RX 9070 XT is still the top GPU. This Gigabyte version is priced at $749, which, of course, is considerably higher than the GPU’s $599 launch MSRP.

The other AMD cards in the chart are an Asus model of the 9070 XT, an Asrock 9070, and an Asus 9060 XT. Nvidia also holds four positions with its RTX 5070, RTX 5060 Ti, and two RTX 5080 models.

Things appear even more favorable for Lisa Su’s firm in Germany. The figures, which come from Mindfactory, show AMD accounting for almost 56% of all GPUs sold over the last week, while Nvidia holds a 40% share. Again, the RX 9070 XT is the most popular card by far, followed by the RX 9060.

It’s worth remembering that Mindfactory has a reputation for favoring AMD, offering promotions, discounts, and bundles featuring its products, which help push up sales figures. The store’s AMD cards have outsold Nvidia’s in the past.

However, these charts do illustrate the impact of the memory crisis on the graphics card market. All GPUs are more expensive, but AMD’s are often cheaper than equivalent Nvidia cards.

The other reason is likely simple availability. The Amazon chart shows three AMD cards in stock, while Nvidia only has one – a Gigabyte RTX 5080 for a colossal $1,599, and that’s with a 6% discount.

Only one company dominates the Steam survey GPU chart

TechEpiphany’s claim of AMD “dominating everywhere” might be a bit of an exaggeration, but there are signs of consumers opting for its cards because of the cheaper prices – or just buying them because there are no Nvidia alternatives. The most valuable company in the world still has GPUs in almost 73% of Steam survey participants’ machines, compared to AMD’s 18.6%, so we’re not seeing a red revolution just yet.