M5 Ultra: Apple’s Most Powerful Chip Yet

The M5 Ultra represents a significant leap in Apple silicon engineering. It features an up-to-36-core CPU with 12 super cores and 24 performance cores, delivering up to 1.3 times higher multithreaded performance than the previous M3 Ultra . The GPU offers up to 80 cores, making it the most powerful graphics processor Apple has ever built, with graphics performance up to 1.8 times faster than the prior generation.

For the first time, Apple has integrated Neural Accelerators directly into every GPU core on an Ultra-class chip . This architectural change delivers up to 4.3 times the peak AI compute performance compared to the M3 Ultra, and an extraordinary 9.8 times improvement over the M1 Ultra . Combined with up to 512GB of unified memory and 1.2TB/s of memory bandwidth—a 50 percent increase over the previous generation—the new Mac Studio can run massive large language models with hundreds of billions of parameters entirely on-device.

The chip is built by connecting two M5 Max dies using Apple’s UltraFusion technology, which increases inter-die bandwidth to over 4.4TB/s and connection density by more than six times, allowing the four dies to function as a single processor.

Breakthrough Storage Architecture

One of the most notable technical advancements in the new Mac Studio is its storage system. According to Apple’s newsroom in multiple language versions—including Chinese, French, German, and Italian—the new Mac Studio features a next-generation SSD architecture built on PCIe Gen 6 . This delivers storage performance up to twice as fast as before, with industry-leading read and write speeds for rapid project loading, file transfers, and loading huge LLMs.

If confirmed, this would make the new Mac Studio the first consumer device to support the PCIe Gen 6 standard . Interestingly, Apple’s English and Spanish language press releases do not explicitly mention “PCIe Gen 6,” though the Italian and Chinese versions do, and IT之家 has reported that Apple is aware of this inconsistency.

AI Performance That Redefines Professional Computing

Apple is positioning the new Mac Studio as the ultimate machine for on-device AI workloads. The numbers speak for themselves:

• CopyCat ML training in Foundry Nuke: Up to 15.4 times faster than M1 Ultra, and 3.3 times faster than M3 Ultra
• LLM prompt processing in LM Studio: Up to 9.8 times faster than M1 Ultra, and 4 times faster than M3 Ultra
• Text-to-image generation: Up to 8.2 times faster than M1 Ultra, and 4.3 times faster than M3 Ultra
• Scene rendering in Maxon Redshift: Up to 4.7 times faster than M1 Ultra, and 1.7 times faster than M3 Ultra

For enterprise teams, multiple Mac Studio systems can be clustered using Thunderbolt 5 and RDMA (remote direct memory access) technology, creating a shared memory pool across systems. Apple says a cluster of four Mac Studio systems delivers up to three times faster AI inference than a single system.

Pro Connectivity and Wireless Upgrades

The new Mac Studio also gains significant connectivity upgrades. Thunderbolt 5 ports deliver transfer speeds up to 120Gb/s, supporting high-performance peripherals, displays, PCIe expansion chassis, and external storage . The Apple-designed N1 chip brings Wi-Fi 7 and Bluetooth 6 to Mac Studio for the first time, delivering improved performance and reliability to wireless connections.

The machine also supports genlock over USB-C for precise synchronization between displays and professional cameras like the iPhone 17 Pro, and can drive up to eight displays or four Studio Display XDRs at full 5K resolution and 120Hz.

Pricing and Availability

The new Mac Studio is available for pre-order now and will begin shipping on September 22, 2026 . Pricing starts at $2,499 for the M5 Max configuration and $5,499 for the M5 Ultra configuration. The fully loaded model with 512GB of unified memory and the 80-core GPU will not ship until late October, with pricing expected to exceed $10,000.

The timing carries special significance. Cook will officially step down as CEO on September 1, handing the reins to hardware engineering chief John Ternus while transitioning to the role of executive chairman. The Mac mini refresh is expected to be the final product released under Cook’s direct leadership.

M5 and M6 Chips Being Tested for the New Desktop

Apple has been testing both M5 and M6 processor versions for the new Mac mini. The update comes as demand for the current Mac mini has surged, driven largely by its popularity as a local platform for running AI applications. However, memory chip shortages are making it difficult for Apple to maintain adequate inventory.

First Mac Mini to Be Assembled in the United States

Beyond the processor upgrade, the new Mac mini represents a milestone in Apple’s U.S. manufacturing push. The company is preparing to begin production of the Mac mini at a Foxconn facility in Houston, marking the first time this model will be assembled in the United States. Apple first announced its Houston manufacturing plans in February, and the same facility is already producing Apple’s AI servers.

What Else Is Coming in the Mac Lineup

Bloomberg’s Mark Gurman reports that additional Mac updates are on the horizon, though they may take longer to arrive. These include a new iMac with fresh color options, also featuring the M6 chip, an upgraded entry-level MacBook Pro, and Apple’s first touchscreen computers, expected in redesigned 14- and 16-inch MacBook Pro models with OLED displays.

A Week of Transition as Cook Prepares to Step Down

The Mac mini launch coincides with what is shaping up to be a historic week for Apple. On August 24, the company held a farewell party for Cook at Apple Park, with about 200 employees in attendance. Speakers included Laurene Powell Jobs (widow of Apple co-founder Steve Jobs), former COO Jeff Williams, services chief Eddy Cue, and incoming CEO John Ternus.

Cook is expected to remain involved as executive chairman, focusing on policy engagement with global decision-makers. For now, company insiders describe the leadership transition with one word: continuity. The new Mac mini, likely arriving within days, will be one of the last products to bear the “designed under Tim Cook” stamp—and the first Mac mini built on American soil.

The move follows a July 18 price hike in Japan, where Apple’s first broad iPhone price adjustment in years saw the top-storage iPhone 17 Pro Max rise by 25,000 yen to 354,800 yen, and the base iPhone 17 climb 13,000 yen to 142,800 yen. Japan’s adjustments ranged from 8 to 11.3 percent, with the iPhone Air seeing the steepest increase at 11.3 percent.

Memory Now Accounts for 34% of iPhone Component Costs

The driver is almost entirely memory. A year ago, memory made up about 10 percent of an iPhone’s component cost; in TrendForce’s third-quarter estimate, that has risen to around 34 percent, and the firm projects it will pass 40 percent in the first half of 2027. Memory prices have climbed five- to sevenfold since the start of 2025, even as panel and application processor costs — historically the two biggest line items — have shrunk.

The buildout of AI data centers has swallowed so much of the world’s memory production that hyperscalers are outbidding every consumer electronics maker for the same wafers. Micron, Samsung, and SK Hynix — the three dominant suppliers — are allocating output to data centers, where margins are better. Data shows mobile DRAM prices have more than doubled since the second quarter of 2025.

“Once-in-a-Century Flood”

Apple CEO Tim Cook described the situation as a “once-in-a-century flood” during the company’s fiscal third-quarter earnings call, noting that prices have risen exponentially. “Unfortunately, price increases are unavoidable. We have done everything we can to absorb cost pressures and avoid passing them on to consumers, but this is no longer sustainable,” Cook told The Wall Street Journal in June.

Cook said Apple is “reluctantly” passing on the costs, adding that DRAM supply is controlled by just three suppliers and that more vendors would help improve pricing.

Washington Opposes Apple’s Search for Cheaper Memory

Apple’s search for cheaper memory also ran into a geopolitical wall. The company had been testing CXMT dynamic random-access memory for devices sold in China, but Commerce Secretary Howard Lutnick expressed the Trump administration’s opposition to the plan.

“The Trump administration is not in favor of that,” Lutnick said after touring an Apple manufacturing facility in Houston. “There have to be other solutions to the memory issue, but it’s not great American companies using Chinese memory.” When asked whether he had relayed this directly to Apple, Lutnick replied he had done so “plainly”.

CXMT and YMTC, the two Chinese memory makers Apple was testing, are both on a Pentagon list of companies linked to China’s military. Apple COO Sabih Khan noted that while Apple customizes many iPhone components, “there isn’t a lot of customization” when it comes to memory, potentially allowing the company to use off-the-shelf Chinese chips.

However, seven U.S. senators, led by Indiana Republican Jim Banks and Senate Democratic Leader Chuck Schumer, gave Apple until August 21 to commit that no CXMT or YMTC memory goes into any Apple product.

What’s Next for iPhone Pricing

The iPhone 17 price adjustment comes as Apple prepares for the September launch of the iPhone 18 lineup. TrendForce estimates the bill of materials for a 256GB iPhone 18 Pro runs roughly 38 percent above the equivalent iPhone 17 Pro, suggesting the next flagship could be $100 to $300 more expensive than the $1,099 iPhone 17 Pro.

Bloomberg’s Mark Gurman wrote in early August that an iPhone price increase was certain, citing memory chip shortages, processor shortages, and the cost of the iPhone 18 Pro’s new camera system. The first foldable iPhone Ultra is expected to start at least $2,000.

For consumers still considering an iPhone 17 purchase, the window may be closing quickly. Apple has not yet officially announced the global price adjustment, but the company already raised Mac and iPad prices worldwide in June, with increases of $100 to $300 across the product line.

A Shift in Content Licensing Strategy

Unlike the fixed-fee model commonly used in the AI industry—such as OpenAI’s arrangements with News Corp or Axel Springer—Apple has proposed a variable compensation plan. Under this framework, publishers would receive payments when their content is actively used by Siri to answer user queries, rather than receiving a blanket upfront fee for broad content access.

People familiar with the discussions say Apple has discussed a total budget that could reach nine figures to fund these payments. The proposed multiyear agreements would grant Apple rights to publisher content for Siri AI features expected to launch later this year.

The Challenge of AI-Generated News

Apple’s current move follows a previous setback in AI-powered news delivery. In late 2024, the company tested an automated feature that generated AI news summaries, but quickly shut it down after it produced inaccurate headlines that raised concerns among publishing partners. This time, Apple appears to be taking a different approach—securing direct access to verified content from publishers rather than relying on automated aggregation.

Broader Context: Publishers Between Licensing and Litigation

The negotiations come at a pivotal moment in the relationship between AI companies and content creators. Publishers increasingly face a choice between licensing content to AI developers or pursuing legal action over unauthorized use of their work. The New York Times is currently suing OpenAI and Microsoft over alleged copyright infringement, while News Corp and Axel Springer have struck content deals with the ChatGPT maker. Reddit has taken both approaches, licensing content to Google for approximately $60 million per year while separately suing Anthropic and Perplexity over alleged unauthorized data scraping.

Apple’s variable-payment model—if successful—could establish a new industry standard. It shifts more financial risk to publishers compared to fixed-fee deals, but could reward outlets whose content proves most valuable to Siri’s underlying AI system.

Not Apple’s First Publisher Partnership

Apple has previous experience with publisher agreements. Since 2019, the company has operated Apple News+, a subscription service that has generated significant revenue for some news and magazine publishers. The Wall Street Journal, whose parent company News Corp has a commercial agreement with Apple, was among the early participants. Apple has also previously signed deals that included rights to use publisher content for AI training purposes.

Why This Matters for Siri

Apple’s timing reflects the competitive pressure in the AI assistant market. Siri has faced years of criticism for limited capabilities compared to rivals like OpenAI’s ChatGPT, Google’s Gemini, and Amazon’s Alexa. The upgraded Siri, expected to launch this fall alongside iOS 27, iPadOS 27, and macOS 27, represents a significant overhaul designed to embed AI more deeply into Apple’s ecosystem. Apple declined to comment when the WSJ reached out for confirmation.

If the deals move forward, they could create a new revenue stream for publishers while giving Apple a competitive edge in delivering accurate, up-to-date information through Siri—addressing a key weakness that has long held the assistant back.

What Remains of the M6 Lineup

The standard M6 chip is still on track for a late 2026 release. It will appear in entry-level MacBook Pro models, Mac mini, iMac, and select iPad Pro and iPad Air configurations.  But that is where the M6 story ends. No Pro, no Max, no Ultra—something that has become a predictable rhythm for Apple’s chip releases is now broken. 

AI as the Driving Force

The abrupt shift comes down to one word: AI. According to Bloomberg’s Mark Gurman, Apple had planned substantial neural-processing upgrades for the M7 family and ultimately decided those improvements were important enough to justify accelerating the next generation rather than completing the M6 product line. 

As Gurman put it in his Power On newsletter, “AI is no longer just a feature that Apple chips need to support—it is now determining how these products are designed and when they are released.” 

The New M7 Roadmap

The accelerated timeline for M7 is equally striking. Apple began taping out the M7 just six months after finalizing the M6 design, compressing what is normally a 12-to-18-month gap into half that time.  The new schedule looks like this: base M7 in the first half of 2027, M7 Pro and M7 Max in the second half of 2027, and the flagship M7 Ultra arriving in 2028. 

Performance targets for the M7 base chip include memory bandwidth of 240 GB/s—roughly 57 percent higher than the M5 Max’s 153 GB/s.  The M7 Ultra is even more ambitious, designed to support up to 1.5 terabytes of unified memory, double the planned capacity of the M5 Ultra.  Apple is reportedly aiming for AI performance that approaches dedicated accelerators like Nvidia’s Blackwell GPUs. 

Implications for Professional Users

The decision creates a significant upgrade gap for high-end Mac users. The redesigned OLED MacBook Pro, originally expected in late 2026, has been pushed to early 2027—and will still ship with M5 Pro and M5 Max processors, offering no performance advantage over current machines. 

Users who need the highest performance will have to wait until late 2027 or 2028 for a true next-generation chip. Apple is attempting to bridge this gap by releasing an M5 Ultra-powered Mac Studio later this year, offering some upgrade path for professionals. 

AI Ambitions Beyond the Mac

The M7 Ultra is not just for Macs. Gurman reports that Apple is evaluating the chip as the backbone for Apple Intelligence servers starting in 2029.  This signals a broader ambition: Apple is designing its Mac silicon to rival dedicated AI infrastructure hardware from Nvidia, a shift that moves the company into an entirely different competitive space. 

The decision to skip the M6 high-end chips was reportedly influenced by additional factors: a global memory shortage that has driven DDR5 prices up by over 300 percent year-over-year, and tightening capacity at TSMC’s 2nm fabs, which are heavily booked by AI companies.  Both pressures made it more practical to conserve resources for the AI-focused M7.

A Strategic Pivot

This is not a minor adjustment. Apple is breaking its own product cadence to chase the AI wave with urgency. The company that once revolutionized mobile computing is now reshaping its silicon strategy around a single priority: making AI run faster, and run locally, on its own hardware.

The adjustments, confirmed by the Financial Times, reflect a growing challenge facing the tech industry: AI systems are now widely used to scan code for potential security issues, but the reports they produce often contain unverified claims or outright “hallucinations” . Apple’s security review team has been overwhelmed by the volume of these submissions, which consume substantial resources and risk obscuring critical vulnerabilities discovered by human researchers.

Security professionals have reported that AI-powered scanning tools can flag thousands of potential issues in a single codebase, many of which turn out to be false positives or misinterpretations of security boundaries. The sheer volume makes it difficult for Apple’s security team to prioritize genuine threats. By capping submissions and requiring researchers to apply for exceptions, Apple aims to restore focus to high-quality, manually validated security research .

Not an Isolated Move

Apple is not alone in responding to this new reality. Google made similar adjustments to its own bug bounty program earlier this year, shifting rewards toward solutions for complex security challenges rather than simple vulnerabilities that AI can easily identify . The move reflects a broader industry recalibration as AI becomes a double-edged sword in cybersecurity—a powerful tool for defenders but also a source of noise that can overwhelm security operations.

For Apple, the timing is notable. The company is in the midst of a major AI push with the launch of Siri AI and Apple Intelligence across its platforms, making the security of its own AI systems an increasingly critical concern . The bug bounty program changes suggest Apple is taking steps to ensure that its security review process remains effective even as the volume of AI-aided research grows.

The new rules apply to submissions made through Apple’s internal security portal, which researchers can access by applying to participate in the bug bounty program. Those who need to submit more than the capped number of reports must now make a formal request to Apple’s security team. The 30-day cooldown period gives researchers time to thoroughly verify their findings before submission, theoretically reducing the number of duplicate or invalid reports.

A Sign of Things to Come

As AI tools for code analysis become more sophisticated and accessible, other tech companies are likely to follow Apple and Google’s lead in adjusting their vulnerability disclosure programs. The challenge lies in balancing the benefits of AI-assisted security research—which can identify patterns and edge cases that humans might miss—against the operational burden of processing machine-generated reports.

Apple’s move also highlights a subtle but important dynamic: while AI is making security research more accessible, it is also creating a quality control problem that the industry must address. The 30-day cooldown and submission caps are essentially friction mechanisms designed to encourage researchers to apply human judgment before submitting reports.

For now, Apple’s bug bounty program remains open to security researchers who find genuine vulnerabilities in Apple software and services. But the new rules make it clear: quality matters more than quantity, and AI-generated noise will not be tolerated.

2nm Process: A Generational Leap

The A20 Pro will be Apple’s first 2nm iPhone chip, transitioning from the 3nm process used in the A19 and A19 Pro . In the semiconductor industry, shrinking the manufacturing process allows more transistors to be packed into the same silicon area. This translates directly to higher processing power at the same power consumption, or significantly extended battery life while maintaining current performance levels .

Apple’s history shows that each manufacturing process transition has delivered measurable gains. The 2nm shift creates additional “space” for Apple to upgrade the CPU, GPU, Neural Engine, and dedicated AI processing components—particularly important as Apple Intelligence features become central to the iPhone experience .

WMCM Packaging: A Structural Breakthrough

Perhaps the more significant change is Apple’s shift from InFO (Integrated Fan-Out) to WMCM (Wafer-level Multi-Chip Module) packaging on the A20 Pro . This marks the first time Apple is using this packaging approach on an iPhone processor.

The technical distinction is important. InFO integrates all components onto a single die, while WMCM combines multiple independent dies—CPU, GPU, Neural Engine—onto a single package at the wafer level before cutting them into individual chips .

This structural change brings several advantages. First, it eliminates the need for interposers and substrates in signal transmission, improving both thermal performance and signal stability . Second, it physically shortens the distance between the processor and memory—a critical factor for AI models running directly on the device, which require extremely high memory access speeds .

Third, WMCM offers greater design flexibility. Apple can mix and match different CPU and GPU configurations, potentially using the A20 series as a foundation to develop M6, M6 Pro, and M6 Max chips for Macs . Fourth, the packaging uses MUF (Molding Underfill) technology, which reduces material consumption and could offset some of the higher costs associated with 2nm production .

Cache and GPU Upgrades

The A20 Pro is expected to receive significant cache upgrades: 16MB L2 cache for performance cores, 8MB L2 cache for efficiency cores, and 36-48MB of system-level cache . The GPU will likely feature third-generation Dynamic Cache, first introduced on the A17 Pro, which allocates on-chip memory in real time based on workload demands—reducing waste and improving performance per watt .

Real-World Impact: AI and Gaming

The combination of 2nm and WMCM packaging is particularly significant for on-device AI processing. When memory sits closer to the processor, data transfers faster and consumes less power—essential for AI models that require high memory access speeds . The A20 Pro’s Neural Engine area is reportedly larger to support these workloads .

High-graphics gaming, 4K and 8K video editing, and real-time image processing also stand to benefit from greater memory bandwidth and lower latency . The improved energy efficiency should help the device sustain high performance without throttling due to heat buildup.

A Strategic Pivot

The A20 Pro’s 2nm process and WMCM packaging represent more than a speed upgrade—they mark a fundamental shift in how Apple designs chips for iPhones . At a time when AI is becoming central to the tech industry, these changes position the iPhone 18 Pro series to maintain its competitive edge against next-generation AI smartphones, while setting a new standard for mobile device performance.

The Leak: What the Documents Reveal

According to the leaked bill of materials, U.S. versions of the iPhone 18 Pro will continue using Qualcomm components, including the Snapdragon X80 5G modem and associated RF chips such as the SDR875, QDM8771, QDM8720, PMK75, PMX75, and QET7100A. International models, by contrast, are expected to ship with Apple’s in-house C2 modem, codenamed Ganymede.

Why Two Modems? The mmWave Problem

The primary reason appears to be mmWave 5G support. Apple’s current C1 and C1X modems do not support mmWave, and the leaked files suggest the C2 will also lack this capability. mmWave is the ultra-high-frequency 5G band used by U.S. carriers like Verizon and AT&T to deliver extremely fast download speeds over short distances. Since those carriers have invested heavily in mmWave infrastructure, they want flagship phones sold in the U.S. to support the technology.

Hardware Differences: Two Logic Boards, Two iPhones

The leaked schematics show Apple is using two different logic board designs. Board number 820-04340-06 includes mmWave connectors and Qualcomm components, while board 820-04305-06 uses the C2 modem and lacks mmWave hardware. This suggests a physical hardware split rather than a simple software configuration difference. The U.S. model is physically different from the international version at the board level.

Battery Life: A Potential Tradeoff for U.S. Users

This dual-modem strategy comes with potential tradeoffs. Apple’s in-house modems are widely regarded as more power efficient than Qualcomm’s solutions, meaning U.S. iPhone 18 Pro buyers could see somewhat worse battery life than those purchasing the same device elsewhere. Daring Fireball’s John Gruber argues that mmWave offers little practical benefit for most users, calling it simply a waste of battery life and suggesting the real driver is carrier bragging points rather than user experience.

eSIM for China and Other Upgrades

The leak also points to other iPhone 18 Pro upgrades, including a new Sony IMX905 primary camera sensor. The configuration list in the leaked files shows that from V64 P2 version onwards, there will be no support for dual physical SIMs, with the CN configuration indicating a combination of eSIM and physical SIM. This means the China market version is likely to introduce eSIM support for the first time on the iPhone 18 Pro.

Apple’s Long Road to Modem Independence

Apple has been working to reduce its reliance on Qualcomm since acquiring Intel’s modem business for $1 billion in 2019. The company’s first in-house C1 modem debuted on the iPhone 16e in early 2025 and has since appeared in the iPhone Air and iPhone 17e. However, flagship models including the iPhone 17, 17 Pro, and 17 Pro Max still use Qualcomm chips. The iPhone 18 Pro would represent a significant expansion of Apple’s modem strategy, even if it falls short of a complete transition away from Qualcomm.

Apple released the first public beta of iOS 27 on July 13, opening the doors for any registered user to try the company’s newest operating system update ahead of its official September launch. After spending several days with the beta, one thing is clear: this update is all about making your iPhone work better, not necessarily look different.

Performance: The Real Star of the Show

If iOS 26 was about aesthetics and Liquid Glass, iOS 27 is about sheer speed. Apple claims over 30 specific performance enhancements across the system, and the numbers speak for themselves. App launch speeds improved by up to 30 percent compared to iOS 26.4.2. Photo loading is now up to 70 percent faster, making browsing your gallery a much snappier experience. AirDrop transfers are 80 percent quicker, with near-instant discovery of nearby devices. These are not just theoretical numbers. On devices ranging from the iPhone 16 Pro down to the five-year-old iPhone 11 Pro Max, the difference is immediately noticeable. Notifications drop down with zero stutter, app switching feels fluid, and the entire interface feels more responsive. As one reviewer put it, the experience has returned to the smoothness of the iOS 18 era. For users who were frustrated by the occasional lag in iOS 26, this is the upgrade you have been waiting for. The system now feels like it is finally catching up to the hardware.

UI Refinements: Listening to Feedback

Apple has also taken user feedback to heart on the controversial Liquid Glass design language. What was once a binary toggle is now a slider, allowing users to adjust the transparency effect from subtle to pronounced. The update also introduces new extra-large widgets that can span nearly the entire screen, offering a more immersive way to interact with apps like Apple Music and Weather. And in a move that has been widely praised, Chinese users get an exclusive feature: holiday alarm scheduling that automatically adjusts for public holidays and workdays.

Siri AI: The Unfulfilled Promise

The biggest disappointment is Siri AI. While the international version gets a major overhaul with contextual awareness, cross-app actions, and a dedicated interface, it remains unavailable in China due to regulatory requirements. Even where it works, the experience is inconsistent. Voice recognition can fail randomly, multi-step tasks often stop after the first action, and answers to real-time questions are sometimes outdated. The feature feels like a work in progress, not the polished, life-changing assistant Apple promised.

Should You Upgrade?

The iOS 27 public beta is surprisingly stable for a beta release. Tech journalist Joanna Stern called it the first beta in years that she would recommend on day one. However, it is still testing software. Users have reported occasional heating issues, battery drain, and app incompatibilities. It is best installed on a secondary device. The final version of iOS 27 is expected to arrive alongside new iPhones in mid-September. For most users, waiting for that polished, stable release will be the smarter choice. But if you value pure performance over new features, iOS 27 is shaping up to be one of Apple’s most impactful updates in years, even if its AI is not quite ready to impress.