It’s spending billions on businesses that could turn financial services into its next growth engine

Grab has spent roughly US$2.5 billion (S$3.2 billion) on acquisitions so far this year. Strip out its expansion into Taiwan, and most of that money is going toward one thing: financial services, particularly lending.

Until recently, Grab was largely building its financial services business from the ground up.

That is changing. This year, it has spent hundreds of millions of dollars buying businesses that already have customers and lending operations, with Atome accounting for the biggest bet so far.

Here’s a look at Grab’s acquisitions, how much it has spent, and what it’s trying to build with those deals.

Grab’s 2026 acquisitions

Grab’s acquisition spree started in Jan, when it invested an undisclosed amount in Chinese robotics startup Infermove.

But the first major deal came a month later, and it was in financial services.

Feb: Stash (US$425 million)

Grab announced in Feb that it would acquire US-based investing app Stash at an enterprise value of US$425 million (S$543 million).

The deal gives Grab access to a digital investing platform with more than US$5 billion in assets under management and over one million paying subscribers. Stash is also expected to generate more than US$60 million in adjusted EBITDA in 2028.

Grab will initially acquire 50.1% of Stash, with the remaining stake to be paid for at fair market value over the following three years.

Mar: foodpanda Taiwan (US$600 million)

The following month, Grab announced a US$600 million (S$767 million) deal to acquire Delivery Hero’s foodpanda delivery business in Taiwan.

This one was different from the Stash deal. Rather than expanding Grab’s financial services business, it was a geographic expansion: Taiwan would become Grab’s ninth market and its first outside Southeast Asia.

The Taiwan business generated about US$1.8 billion in gross merchandise value in 2025 and was profitable on an adjusted EBITDA basis. Grab plans to migrate foodpanda’s users, merchants and delivery partners onto its platform, with the full migration targeted for early 2027.

Sept: Atome Financial (US$1.49 billion)

Then came Grab’s biggest deal of the year so far.

In Sept, Grab agreed to pay US$1.49 billion (S$1.91 billion) for a 60% stake in Atome Financial, the digital lending business of Advance Intelligence Group.

Atome operates across Singapore, Malaysia, the Philippines, Indonesia and Thailand, offering Buy Now, Pay Later loans, consumer cash loans, BNPL cards and other digital lending products. It has 25 million cumulative transacted users and a gross loan portfolio of about US$1 billion.

The remaining 40% will be acquired roughly two years after the first transaction closes. Unlike the initial 60%, that portion does not have a fixed price: its valuation will depend on Atome’s revenue and adjusted EBITDA, subject to a floor of US$2 billion and a cap of US$4.5 billion.

That makes Atome by far the biggest financial services bet in Grab’s 2026 acquisition spree—and the deal that most clearly shows where Grab is putting its money.

The bet on lending

The common thread between Stash and Atome is that Grab is not simply buying another app or adding another product to its super app.

It is buying established financial businesses with existing customers, products and lending infrastructure—giving it a faster route to scale its financial services operations than building everything in-house.

But why now?

The strategy comes as Grab’s financial services business begins to outpace its more established mobility and delivery businesses.

In Q2 2026, financial services revenue grew 59% year-on-year to US$134 million, compared with 21% for deliveries and 12% for mobility. Lending has also accelerated, with total loans disbursed rising 72% to a record US$1.2 billion.

Part of the sharp increase in Grab’s loan book came from Superbank, an Indonesian digital bank that Grab began consolidating in Jun after increasing its stake to more than 50%. Grab had been an investor in Superbank since 2022, and the bank had more than six million customers by April 2026.

Grab’s gross loan portfolio reached US$2.3 billion at the end of Q2, up 197% from US$781 million a year earlier. Superbank contributed US$761 million of loan receivables to the portfolio, although Grab said the loan book would still have doubled year-on-year even without its contribution.

The business is not profitable yet, however.

Financial Services recorded an adjusted EBITDA loss of US$15 million in Q2, although that was an improvement from the US$26 million loss a year earlier. Grab attributed the improvement to higher contributions from its lending businesses and Superbank.

The company is betting that more scale will change that equation.

With Atome included, Grab expects its Financial Services segment to reach US$500 million in adjusted EBITDA and a gross loan portfolio of more than US$6 billion by 2028.

At the same time, Grab’s core business has finally started generating meaningful cash.

Grab reported its first full-year profit in 2025, and by the end of Q2 2026 it had US$7.4 billion in cash and cash equivalents.

That gives Grab room to spend on acquisitions while continuing to return money to shareholders.

Becoming the “everything” app

Grab has spent over a decade trying to become the “everything app” of Southeast Asia—the one platform a person opens for a ride, a meal, a parcel, and, eventually, their money.

Ride-hailing may have been the entry point, but financial services are increasingly becoming a bigger part of that strategy. Grab already has payments, digital banking and lending products. With Stash and Atome, it is adding investing and a much larger consumer lending business to that stack.

But the bigger question is whether Grab can turn that scale into sustainable profits.

The company now has a much larger lending operation, millions of additional financial services customers and a broader range of products. But it also has greater exposure to credit losses as it lends to more consumers across the region.

That makes the next few years particularly important. Grab has the cash to keep making acquisitions, but its US$2.5 billion spending spree will ultimately have to show up in the numbers—especially if financial services is to become the profit engine management expects it to be by 2028.

• Read other articles we’ve written on Singaporean businesses here.

James Channel recently got hold of a small box that promised the latest Windows system, complete with a keyboard, mouse, and enough software to turn any living room into a home office. What arrived instead was a Famiclone, one of the countless unauthorized copies of Nintendo’s original Family Computer that flooded markets across Asia and beyond during the late 1980s and 1990s. These machines used a single chip to recreate the hardware that once powered Super Mario Bros., and manufacturers kept producing them long after the patents expired because they were cheap to build and easy to sell to families who could not afford a real computer.

The packaging for this particular machine really pushed Windows, as the box art was a mishmash of screens that looked like classic desktop software if you squinted hard enough, and the text advertising all sorts of features, like a light gun and multitasking, sounded a little suspect. The unit was housed in a plastic shell the size of a decent thick paperback, and if it had been a little more substantial, it might have had a chance, along with two joystick ports that appeared to have been lifted directly from a Commodore 64, some audio and video cables, a power adapter with a load of Japanese text on it, and a pair of cartridges. One of them was this 148-in-1 educational computer label, oh right, and the other boasted an outrageous number of games, which showed out to be approximately 140 with more than a few duplicates. A completely legitimate Indonesian version of a comparable cartridge appeared as well, however it was connected incorrectly and loaded with local language lessons.

PlayStation®5 Digital Edition – 825GB

• Slim Design, players get powerful gaming technology packed inside a sleek and compact console design.
• 825GB of storage, keep your favorite games raeady and waiting for you to jump in and play
• Ultra-High Speed SSD, maximize yoru play sessions with near instant load times for installed PS5 games

The keyboard, however, was a completely different story, since it was the worst piece of equipment he’d ever seen from the era. The keys were somewhat crooked, and the space bar was misspelled; come on, how hard is it to get that right? The membrane underneath seemed as if it would crack at the least provocation, so they had to go ahead and stick the Windows keys on there, but they didn’t work. Out of all of this, the mouse, a classic ball type with a star-shaped roller thingy inside, was a bit of a joke, as the ball kept falling out and the x-axis refused to register until I had to use super glue and a blowtorch to keep everything together. What about the controller that came with the set? Well, the cables were really thin, and everything felt stupidly loose in my palm. Apparently, the gun-shaped peripheral that came with the package was thrown in the dumpster by Australian customs because it looked suspiciously like an airgun, go figure.

The cartridges themselves were jam-packed with a combination of borrowed NES games and simple educational programs for children. GBasic appeared, and it looked strangely like Nintendo’s old Family Basic cartridge, allowing users to enter in small little programs that vanished as soon as you switched it off. Minesweeper and Solitaire ran in a style that recalled Windows 3.1, complete with a chording system that allowed you to click on cells and acquire additional tiles by pressing two mouse buttons at the same time. The typing lessons were also a bit of a joke, with fingers moving along the keyboard at about 30 characters per minute, and the other titles included a basic calculator, a music player that sounded like it had been tuned by a novice, a painter that only responded to keyboard input, and an English-Indonesian dictionary, because all you really need is the ability to look up Bahasa in the middle of a game.

Opening the case revealed how little money was actually invested in this build. There was a tiny little voltage regulator simply chillin’ by the power input, and one solitary transistor attempting to handle the visual amplification, albeit it was probably straining a little, considering that many other components had been left off the board to save money. The RF connector no longer modulated the signal; instead, it was bridged over, allowing audio to pass through without modification. Then there was a separate chip, some DH420D, that accepted keyboard pushes and converted them into the signals that a Famicom expects, all thanks to Sujia, a Chinese manufacturer that also branded a slew of generic boards. Some wires’ insulation had even melted in areas due to their proximity to the shell. To make matters worse, the power switch contacts were so blocked with grime that you had to give them a thorough cleaning and lubricant just to get the device to turn on properly.

Software-wise, it was a bit of a mess, since the restricted memory would load up and then freeze on you, forcing you to reset it and wiping out all of your work in the process. You could miraculously get some NES games to operate after unplugging the knackered old mouse. To be honest, the keyboard controls were somewhat of a joke for platformers. The educational portions of the software served as basic drills, but the music was out of sync, and the frame rate would plummet when things got too difficult. These machines were marketed to households where proper computers were out of reach due to their high cost, therefore they were an acceptable way to gain experience with technology that cost a fraction of the price of a real PC. Even though they don’t quite live up to what the box promised, collectors continue to seek them out because of the unusual combination of a full keyboard, a working mouse, and cartridge-based lessons, which is a fairly uncommon find among Famiclones, even if the end result isn’t exactly smooth.

Record labels spent the better part of two decades convincing consumers that owning music was increasingly unnecessary. Now that vinyl and CDs are enjoying a genuine resurgence, the same industry seems determined to find out just how much collectors are willing to spend on albums they probably already own.

Apparently, $300 is no longer out of the question.

On October 9, Rhino High Fidelity released a limited reel-to-reel edition of Joni Mitchell’s Court and Spark for $299.98. Only 350 copies were offered worldwide, and by Saturday morning, Rhino’s website was already showing the title as sold out. That’s an impressive turnaround for an album that requires playback equipment most consumers have never seen outside a recording studio.

On the very same day, Verve announced a $124.98 One Step vinyl edition of Diana Krall’s The Look of Love, celebrating the album’s 25th anniversary. The November 20 release joins a growing collection of premium audiophile records that now cost more than many people spend on music in an entire year.

Both releases are aimed at serious collectors, but they raise a larger question. Is the physical media revival encouraging record labels to produce genuinely better editions, or have they discovered that scarcity, elaborate packaging, and audiophile marketing can make almost any price seem reasonable?

Related Reading:

Physical Music Is Growing and the Labels Have Noticed

The numbers explain why record companies are investing so heavily in physical formats.

According to the Recording Industry Association of America (RIAA), U.S. physical music revenue increased 25.9% during the first half of 2026, reaching $731.5 million. Vinyl revenue climbed 17.7% to $543.8 million, while CD revenue surged 58.6% to $171.1 million.

As we examined in our recent report on the CD resurgence, the rebound is not limited to aging audiophiles rediscovering their turntables. Younger consumers, collectors, and music fans who grew up with streaming are also buying physical releases.

Streaming still dominates the recorded music business, accounting for approximately 82% of U.S. revenue during the first half of 2026. Physical music is not threatening that position, but it has become a valuable complementary market built around ownership, better packaging, and the desire to support artists.

Record labels have figured out something rather important: consumers who already have access to virtually everything are increasingly willing to pay for something they can actually keep.

The danger is assuming that willingness has no financial limit.

Rhino Sells 350 Copies of Joni Mitchell’s Court and Spark Almost Overnight

Rhino deserves some credit here.

Its new Court and Spark reel-to-reel edition is not simply another colored pressing with a numbered sticker and a promise that this time the music will sound even more magical.

The company has created a legitimate specialty product with a manufacturing process that is considerably more demanding than pressing another batch of records.

The tape was duplicated in real time from a 1:1 copy of the original stereo master. It uses quarter-inch RTM LPR90 tape stock, runs at 15 inches per second in half-track stereo, follows the IEC equalization standard, and comes on a 10.5-inch metal reel.

Those details are important. Real-time tape duplication is labor-intensive, the materials are expensive, and production cannot be scaled as easily as conventional vinyl manufacturing. This also isn’t a copy of the actual original master tape; Rhino specifies that the duplication source is a first-generation 1:1 copy.

The result should appeal to an extremely small audience with the hardware, technical knowledge, and financial resources to appreciate what Rhino is offering. A properly configured 15 ips half-track reel-to-reel machine isn’t something you casually pick up at Target alongside laundry detergent.

But Rhino understood its customers.

Limiting the edition to 350 copies worldwide created genuine scarcity, and the fact that the company’s online store listed it as sold out by the following morning suggests that the demand was very real. For a $300 analog release in 2026, that is a remarkable achievement.

The clever part is that Rhino hasn’t made this the only serious way to experience the album.

In July, eCoustics contributor Mark Smotroff reviewed Rhino High Fidelity’s $39.98 all-analog vinyl edition of Court and Spark, mastered by Kevin Gray and pressed at Optimal Media. Mark found it superior to his Nautilus SuperDisc edition, praising its more natural presentation, improved separation, and greater warmth.

That’s a very good record for less than $40.

Collectors with suitable tape machines can spend $300 on the new reel-to-reel edition, while listeners who simply want a superb analog presentation of Mitchell’s 1974 masterpiece have a considerably more affordable option.

Rhino hasn’t proven that $300 tapes are the future of physical music. It has demonstrated that a carefully chosen title, credible production methods, and genuinely limited availability can create a successful niche product.

Mazel tov. Now try finding a properly serviced reel to reel tape machine without disregarding your mortgage and private school bills.

Diana Krall’s $125 One Step Vinyl Raises a Different Question

Diana Krall’s The Look of Love is certainly worthy of a premium reissue.

Released in 2001, The Look of Love found Diana Krall working with producer Tommy LiPuma, legendary arranger Claus Ogerman, and the London Symphony Orchestra. Al Schmitt handled the recording and earned a Grammy for his efforts. Not a bad collection of talent for an album that would become one of Krall’s biggest commercial successes.

And the results speak for themselves. Krall’s smoky vocals, understated piano playing, and Ogerman’s lush orchestral arrangements make this the kind of recording that can sound wonderful on a modest system but really comes alive when the equipment is up to the task. Whether it’s worth spending $125 to hear it on yet another vinyl pressing is an entirely different question.

Verve’s new edition uses the One Step manufacturing process, which eliminates intermediate metal-plating generations between the lacquer and the finished stamper. The objective is to reduce potential losses introduced during manufacturing and preserve more information from the lacquer cut.

That approach can produce excellent results, but it also raises manufacturing costs because fewer pressings can be produced from each lacquer before additional cutting and plating work becomes necessary.

The new two-LP edition will be individually numbered and packaged in a gatefold tip-on jacket with a specially designed slipcase.

All perfectly reasonable.

The $124.98 asking price, however, deserves more scrutiny.

Verve already offers a conventional 2-LP edition of The Look of Love for $34.99 at Amazon. More revealing is the $19.98 deluxe anniversary CD, which includes five bonus tracks, including an alternate take of “Maybe You’ll Be There” and the previously unreleased vault performance “I Love Being Here With You.”

The $125 One Step vinyl contains only the original ten-track album.

In other words, collectors paying more than six times the price of the deluxe CD receive the premium pressing and packaging, but none of the additional music.

Apparently, the bonus tracks didn’t qualify for first-class seating.

There is also the issue of transparency.

Verve’s announcement and current product listing explain the advantages of One Step manufacturing, but do not identify the mastering engineer, specify the source used for the new lacquer cut, or confirm whether the mastering chain is entirely analog.

That information should be available before someone commits $125.

A One Step pressing can be produced from an analog source or a digital source. The manufacturing process alone does not establish which was used, nor does it guarantee that the finished record will sound better than a well-mastered conventional pressing.

This is not a criticism of how the new Krall edition sounds. Nobody should make that judgment without listening to it.

But the burden should be on the label to explain exactly what collectors are getting for their money.

When Does Premium Become Excessive?

We’ve covered enough audiophile reissues in 2026 to know that expensive doesn’t automatically mean better, and affordable doesn’t mean compromised.

Mobile Fidelity’s upcoming Michael Jackson BAD UltraDisc One Step edition also costs $125, but MoFi identifies its source as the album’s original Mitsubishi X-80 16-bit/48kHz digital master, along with the mastering engineer, cutting chain, and pressing facility.

More recently, we examined Beck’s $110 Ride Lonesome One Step release, which uses 24-bit/96kHz digital files transferred from analog masters. That information gives buyers an opportunity to decide whether the production approach and price make sense to them.

Digital mastering isn’t inherently inferior. An expensive record doesn’t become fraudulent because a computer appeared somewhere in its production history. What matters is the quality of the source, the mastering decisions, the pressing, and whether the label has been forthcoming about the process.

Our reviews have also demonstrated that excellent audiophile vinyl doesn’t require a three-figure investment.

Mark Smotroff recently praised Verve’s Acoustic Sounds reissue of Stan Getz and Luiz Bonfá’s Jazz Samba Encore!, a beautifully manufactured all-analog release available for less than $40. He reached a similarly positive conclusion about John Coltrane’s Africa/Brass, another Acoustic Sounds pressing that makes an important recording accessible without requiring collectors to sacrifice their grocery budget.

These are the kinds of releases that have helped establish credibility for the modern audiophile reissue market.

The labels invest in quality mastering, reputable pressing plants, and attractive packaging, while keeping prices within reach of listeners who care more about the music than the resale value of a numbered box.

Nobody is suggesting that a meticulously produced $300 reel-to-reel tape should cost the same as a standard LP. Nor is a $125 One Step pressing automatically a rip-off.

But increasingly expensive collector editions shouldn’t become the default measure of how seriously someone takes music.

The industry also needs to remember that its fastest-growing major physical format in the first half of 2026 was the compact disc, which offers a much more affordable entry point for consumers who want permanent ownership.

There is a message in that statistic, assuming the labels are willing to hear it over the sound of cash registers.

The Bottom Line

Rhino deserves congratulations for selling out its $300 Court and Spark reel-to-reel edition so quickly. The company identified a small audience, produced something technically distinctive, and created enough demand to exhaust a limited run of just 350 copies. It also offers an excellent $40 vinyl alternative for those of us not currently shopping for another tape machine.

Verve’s $125 The Look of Love One Step edition may turn out to be spectacular. The album certainly deserves first-rate treatment, and the company has a strong record of producing outstanding audiophile releases. But buyers should expect more technical transparency when the premium pressing costs several times as much as an existing version of the same recording.

Physical media is thriving because listeners value ownership, permanence, and a closer connection to music. Labels should be encouraged to preserve important recordings, restore overlooked albums, and invest in better manufacturing rather than treating their catalogs as disposable streaming inventory.

What they shouldn’t do is confuse that loyalty with an unlimited willingness to pay.

There’s room in the market for a $20 CD, a superb $40 LP, and even a meticulously produced $300 reel-to-reel tape. Not every format needs to appeal to everyone, and collectors should be free to spend their money however they please.

But if record labels turn the physical media revival into an increasingly expensive competition over numbered editions, extravagant packaging, and questionable value, they risk pricing the next generation of music buyers out of the hobby entirely.

Which would be quite an accomplishment.

After spending twenty years convincing consumers that they didn’t need to own music, the industry may now discover that people still want to own it.

They just might not want to sell their second favourite child to prove it.

Related Reading:

You’ve probably heard of Mars Pathfinder, the NASA lander and its Sojourner rover that took up residence on Mars in 1997. At first, all was well, but after a few days, the lander suffered a series of total system resets. A watchdog timer detected a problem and forced reboots. [Vivek Bhageria’s] recent post looks into what happened and the eventual solution.

The root cause was priority inversion. This happens when a lower-priority task holds a resource needed by a higher-priority task and prevents the higher-priority task from executing. In this case, the operating system was VxWorks, but this can happen on any kind of priority scheduling system.

The simplified version is that there were two tasks that each used the same mutex to avoid stepping on each other. The highest priority was a data distribution task. The lowest priority task dealt with meteorology data processing. There was also a medium-priority task that handled bus maintenance.

The failure would occur when the low-priority task had the mutex and lost its time slot before releasing it. The high-priority task would then find it couldn’t take the mutex. However, the low-priority task didn’t get control again because of the medium-priority tasks waiting to run.

In effect, the high-priority task was now waiting on something that couldn’t become available until the lowest-priority task was able to run. That’s priority inversion.

Sadly, engineers noticed this happening occasionally during testing but, as we are often prone to do, they chalked it up to glitches. As one of our old bosses used to say: “If you don’t know why it happened, then you don’t know that it won’t happen again.”

VxWorks, of course, understands that this might happen and, luckily, there was enough capability on the spacecraft to turn priority inheritance on for the mutex in question. This allows the low-priority task to get a boost because a higher-priority task is waiting on it.

You can read more about the details in the post. We also covered this a decade ago with our own analysis.

Album and playlist pages in YouTube Music now open with cover art that fills the upper third of the screen for some users.

The change surfaced through several Reddit posts from users who received the new layout, though it looks like a server-side A/B test rather than an app update anyone can download.

A server-side switch lets Google enable the design for a small slice of accounts and gauge the response, a rollout method the company regularly uses before pushing interface changes across its wider user base.

Elon Musk’s reach spans multiple global industries, placing the billionaire (sometimes trillionaire, depending on the stock market) at the center of the electric-vehicle industry, space exploration, artificial intelligence and national politics. He’s already involved in telecommunications through his company SpaceX and its global satellite internet service, Starlink, but recent developments may soon expand his influence with satellite-to-cellular coverage.

SpaceX announced on X Thursday that it has agreed to acquire nationwide 800 MHz low-band spectrum to help build Starlink Mobile into a major US cellular carrier. By acquiring these low-band airwaves, SpaceX said it aims to improve signal propagation, allowing satellite coverage to bypass physical obstacles like building walls, dense foliage, and hilly terrain

“This is the last critical piece of the spectrum puzzle needed for SpaceX to provide complete phone coverage in America,” Musk posted on X.

A SpaceX representative did not immediately respond to a request for comment.

Under the terms of Thursday’s announcement, SpaceX will acquire investment firm Grain Management’s nationwide 800 MHz low-band spectrum portfolio for an estimated $8 billion. The deal gives the satellite operator up to 14 megahertz of paired airwaves, enabling two-way data transmission between satellites and mobile devices. The acquisition remains subject to approval by the Federal Communications Commission, which has not yet set a timeline for a decision.

Acquiring the low-band spectrum is crucial for Starlink Mobile’s cellular ambitions. Unlike the high-frequency beams used currently, these low-band airwaves possess the physical properties required to pierce concrete walls, dense foliage and urban infrastructure, effectively solving the indoor dead zone problem that has long limited satellite-to-phone connectivity.

Also this week, the FCC gave SpaceX the go-ahead to launch 15,000 satellites, which the company said can deliver more than 100 times the bandwidth of its current satellites. These satellites are direct to device, meaning that they can transmit phone calls, texts and other data straight from space to devices on the ground. SpaceX said the new satellites will bring “high-speed service directly to unmodified devices anywhere in the world.”

What does it mean for consumers?

For people looking for an alternative to the big three cellphone carriers AT&T, T-Mobile and Verizon Wireless, Starlink Mobile could eventually be an option.

If and when the FCC approves the deal with Grain Management, Starlink Mobile will combine its newly acquired low-band spectrum with a massive, freshly approved fleet of 15,000 next-generation satellites. This hybrid network will use a mix of satellite beams and ground infrastructure to penetrate walls and eliminate indoor dead zones across the US while leveraging its high-bandwidth frequencies to deliver high-speed data directly to ordinary mobile devices. SpaceX plans to accelerate deployment of the new system next year.

But according to analyst Tim Farrar, an industry expert with research firm Telecom, Media and Finance Associates, simply acquiring low-spectrum bandwidth in this week’s deal won’t be enough to compete strongly against AT&T, T-Mobile and Verizon Wireless, which uses thousands of cellphone towers.

“It’s still a very limited amount of spectrum, and to get reliable building penetration in urban areas, SpaceX would have to deploy towers on the ground,” Farrar told CNBC.

Last month, T-Mobile’s chief financial officer, Peter Osvaldik, dismissed the idea that Starlink’s satellite-to-cell service could adequately compete with the traditional wireless coverage offered by the major carriers. “It’s not a competitive threat,” Osvaldik said at the Global TMT Conference.

“By the time the signal has gotten to your phone, you have no signal strength left,” he added. “You can have a million satellites up there with all the spectrum in the world, you can’t break the fundamental laws of physics.”

Ever since being admittedly fascinated by the Cambridge coffee webcam from the 1990s, I’ve written about VPNs, the NFL, smartphones, living wages, over/unders and everything in between.

See full bio

PC shipments saw a steep drop of about 20 percent in Q3 2026, analysts reported this week.

The industry saw its “sharpest decline since “Q1 2023,” Omdia said in an announcement toady. Global shipments of laptops, desktops, and workstations fell 21.2 percent year over year (YoY) to 58.1 million devices, the analyst said.

Desktop PC shipments (including desktop workstations) decreased 23.5 percent to 11.7 million units, and laptop shipments (including laptop workstations) declined 20.6 percent to 46.4 million, per Omdia.

IDC shared similar findings. The analyst said this week that PC shipments fell 20.1 percent YoY to 62.7 million units, compared to 78.5 million in Q3 2025.

Advertisement

The industry also shipped 9.1 percent fewer PCs in Q3 2026 than it did in Q2 2026, when 68.2 million units shipped, per IDC.

The 20.1 percent YoY decline is the biggest drop that the IDC has ever recorded during a year’s third quarter, the IDC said, per PCMag.

The drop in shipments is also notable considering that Q3 PC shipments usually surpass Q2’s numbers.

“If looking at all quarters, then we have seen other quarters in recent years with larger declines, such as 2022 Q4, which was -28.2 percent, and 2023 Q1, which was -28.7 percent. Both were coming off pandemic highs,” Jitesh Ubrani, research director for consumer devices at IDC, told PCMag.

Advertisement

In its announcement this week, IDC said that many PC vendors and sales channels stocked up on PC shipments earlier in the year in anticipation of memory price hikes. That meant minimal demand remained during Q3.

“[Sales] channels are now worried about carrying too much inventory into a market where high prices are suppressing demand,” Ubrani said in a statement.

PDFBear is a broad browser-based PDF toolkit that covers common jobs such as converting, merging, splitting, compressing, signing, and organizing files without installing desktop software. Its standard tools can be useful for occasional PDF work, but the free-plan limits, cloud file handling, Premium-only AI features, and a few unclear product details are worth understanding before you upload important documents or pay.

This is a research-based review, not a hands-on performance test. The verdict therefore focuses on what PDFBear currently offers, what its published policies say, where those policies create practical trade-offs, and which claims cannot be independently confirmed from the available evidence.

Is PDFBear Worth Using?

PDFBear is worth considering for occasional browser-based PDF work because its standard toolset covers many everyday tasks without requiring a paid subscription. The current free plan includes standard conversion, merge, split, compression, and signing tools, with files up to 50 MB.

The main question is not whether PDFBear has enough tools. It does. The more useful question is whether its operating model fits the documents you handle.

If you mainly need to convert a Word document, combine a few PDFs, compress an attachment, rotate pages, or sign a document from a browser, the free tier may cover the job. A registered free account currently supports unlimited conversions with up to 30 pages or images per conversion, while guests are limited to five conversions per day with up to 10 pages or images each.

Premium becomes more relevant when you need optical character recognition (OCR), AI document tools, larger page batches, priority processing, or longer file availability. OCR is the technology that turns text inside scanned pages into searchable and selectable text.

The biggest caution is privacy rather than basic functionality. PDFBear is an online service, so documents are uploaded for processing instead of remaining entirely on your computer. That does not automatically make the service unsuitable, but it matters if the file contains information you would normally avoid sending to a third-party cloud service.

How This PDFBear Review Was Researched

That distinction matters because feature availability does not prove quality. A service can offer PDF-to-Word conversion without that establishing how accurately it reproduces a difficult document. Likewise, PDFBear advertises priority processing for Premium users, but that does not provide an independent benchmark for how much faster it is.

Independent review evidence is also limited. G2 currently shows six PDFBear reviews with a 2.4 out of 5 average. Six reviews can surface individual experiences, but the sample is too small to support a broad conclusion about reliability, support quality, or typical user satisfaction.

What PDFBear Can Do

PDFBear is more than a single converter. Its current tool catalog covers conversion, organization, optimization, editing, signing, security, OCR, and AI-assisted document work.

The standard tools include familiar jobs such as converting Word, Excel, PowerPoint, images, and other formats to PDF; converting PDFs back to common formats; merging and splitting documents; compressing files; rotating or organizing pages; signing PDFs; filling forms; and applying basic document-security tools. PDFBear says these standard tools are available without Premium.

The service also includes more specialized tools such as OCR, Chat with PDF, summarization, translation, and a PDF question generator. These are materially different from a normal file conversion because some depend on artificial intelligence and are reserved for Premium accounts.

That separation is useful when deciding whether you actually need a subscription. Someone who occasionally merges invoices and converts Office files has a very different workload from someone extracting text from hundreds of scanned pages or asking an AI system to summarize long reports.

PDFBear operates through a web browser, so its core tools do not require desktop installation. That cross-device convenience is one of the service’s clearer strengths, especially when you move between computers or only need a PDF tool occasionally.

PDFBear Free vs Premium: Limits and Pricing

The free plan handles routine PDF work, while Premium mainly raises processing allowances and unlocks OCR, AI features, batch processing, priority service, and an ad-free experience. The distinction is easier to understand when the limits are placed side by side.

PDFBear currently offers a 7-day Premium trial and a $99.99 yearly plan. Its FAQ also lists a $13.99 monthly option. Prices and promotional terms can change, so check the checkout page before subscribing.

There is an important qualification behind the word “unlimited.” Standard Premium tasks may have no daily conversion limit, but PDFBear’s AI features remain subject to fair-use controls. The Terms describe per-file page limits, a monthly AI allowance, a maximum upload size of 50 MB per file, and restrictions on automated or unusually high-volume processing.

Premium is therefore better understood as a higher-capacity individual plan than an unrestricted bulk-processing service. If your workflow is automated, shared across many people, or processes unusually large numbers of documents, a normal Premium account should not be assumed to cover it.

Warning
!PDFBear’s official documentation currently conflicts regarding watermarks on free conversions. Its pricing page lists watermark-free PDF output as a Premium-only feature, and its developer documentation states that free conversions receive a small “PDFBEAR.COM – Free Ver.” badge. Conversely, its FAQ states converted files carry no watermark. If clean, unbranded output is essential for client-facing documents, portfolios, or resumes, run a sample test on the exact tool before relying on the free tier.

Because those first-party statements conflict, do not assume one rule applies uniformly to every free website workflow. Testing your specific file path beforehand prevents unexpected formatting revisions when submitting time-sensitive files.

How PDFBear Handles Your Files and Privacy

Uploading a PDF to an online converter means the document leaves your device for processing. For routine files that may be an acceptable trade-off for convenience, but sensitive documents deserve a closer look at storage, deletion, sharing, and AI processing.

Files are transferred through encrypted HTTPS connections while moving between your device and PDFBear. HTTPS protects data in transit, but it is not the same as an independent guarantee that no security incident can occur.

For a free account, a file is removed 30 days after its last activity. Downloading the file or opening its share link counts as activity and resets that timer. Premium files remain available while the subscription is active and are removed 30 days after the subscription ends.

Signed-in users can also delete files manually from the file manager. PDFBear’s Privacy Policy states that uploaded and converted files may be stored on S3-compatible infrastructure such as Wasabi or Amazon Web Services.

Sharing creates another consideration. A shared page is configured so search engines should not index it, but anyone who has the share link can view the file until it is deleted. A no-index setting reduces search-engine discovery; it does not restrict access to one named recipient.

AI tools add another data path. When you use Chat with PDF, summarization, translation, the question generator, or other AI-assisted functions, relevant file content or text may be sent to a third-party AI provider to generate the result. Ordinary conversion should not be described as the same process simply because AI tools are available elsewhere on the service.

If a document contains confidential business information, identity documents, medical information, financial records, private legal material, or another category your organization restricts from external AI services, that distinction matters. Follow the applicable data policy rather than assuming an online AI feature is suitable because the underlying website uses HTTPS.

PDFBear states that human review is not part of its normal automated document-processing workflow. That is a provider statement about its operating practice, not an independent security certification, so it should not be treated as proof that every possible privacy or security risk has been eliminated.

Where PDFBear Falls Short

The first limitation is structural: PDFBear is primarily an online workflow. If your requirement is that a confidential document never leave your own computer, a remote processing service is the wrong model regardless of how convenient its interface may be.

The 50 MB per-file ceiling can also become restrictive with high-resolution scans, image-heavy reports, design files, or long archives. Premium raises the number of pages or images allowed in a conversion, but the current Terms still cap an uploaded file at 50 MB.

Premium’s AI tools also come with limits behind the headline feature list. OCR, Chat with PDF, summarization, translation, and the question generator are included with Premium, but AI usage has monthly and per-file fair-use controls. Heavy AI users should not interpret Premium as unlimited automated document analysis.

There is also no basis for promising flawless output. PDFBear’s Terms state that conversions and AI outputs are not guaranteed to be error-free and advise users to keep backups of important files. That matters most for complex tables, uncommon fonts, scanned text, forms, and intricate layouts, where a converted result should be checked before you rely on it or discard the original.

Free-output watermarking remains unclear because PDFBear’s own current pages disagree. If an unbranded PDF is essential to your workflow, confirm the result with the specific free tool before relying on it repeatedly.

Independent review evidence is too limited for a strong reliability verdict. G2 currently lists only six PDFBear reviews. That is enough to justify reading individual complaints or praise before subscribing, but not enough to establish how the service performs for most users.

PDFBear Alternatives

The most useful alternative depends on why PDFBear does not fit. Replacing it with another cloud converter does not solve much if your actual concern is keeping documents off third-party servers.

If your main task is Merge PDF, compare free PDF merger tools by processing location, free-use restrictions, offline support, and document controls. That comparison is more useful than paying for a broad PDF suite when combining files is the only job you regularly perform.

If you still want another browser-based conversion and editing service, CocoDoc PDF Converter is another option to evaluate. The important comparison is not simply which site has more tools. Check the current free limits, processing model, privacy policy, pricing, and the exact functions you need before moving a sensitive or repeated workflow.

For documents that must stay local, prioritize an offline desktop application or a browser-local tool whose processing genuinely occurs on your own device. Dedicated desktop software such as PDFgear or open-source editors like LibreOffice Draw process files entirely on local hardware without sending document bytes across the network. Choosing a true offline tool or a client-side WebAssembly processor changes the privacy model far more meaningfully than switching between two competing cloud converters.

Who Should and Shouldn’t Use PDFBear

PDFBear is a reasonable fit for people who occasionally need common PDF tasks from a browser and do not want to install dedicated desktop software. The free account is especially relevant when your files stay below 50 MB and your conversions fit within the current 30-page or image allowance.

Premium makes more sense when OCR, AI document tools, batches of up to 200 pages or images, priority processing, or longer file availability are part of your regular work. Even then, the AI fair-use policy should be checked if you expect unusually heavy use.

PDFBear is a weaker fit if you require local-only processing, routinely work with files above 50 MB, need independently benchmarked conversion accuracy, or plan to automate high-volume PDF processing through a normal consumer Premium account.

It also deserves extra scrutiny when the documents are highly sensitive. The service publishes information about HTTPS, retention, deletion, storage providers, and AI subprocessors, which gives you useful policy details to evaluate. Those disclosures still do not remove the basic fact that an online workflow transfers document data to external infrastructure.

If your decision comes down to privacy, start with the data rather than the tool list: identify what the document contains, whether external processing is permitted, whether an AI feature is involved, and how soon you need the result removed. That will usually tell you more than a long feature comparison.

Verdict