EDA & Semiconductor IP

Jim Cramer said selling Arm Holdings early was a "big mistake," noting on the September 16 episode of Mad Money that the stock has fallen 46% from its June high of $452 to just under $244. Arm's fiscal first quarter 2027 results showed total revenue of $1.29 billion, up 22% year over year, with royalty revenue up 22% to $715 million and licensing revenue up 23% to $574 million, while data-center royalties more than doubled and the company posted a non-GAAP operating margin of 41% and $665 million in non-GAAP free cash flow. Management expects Q2 FY27 revenue of $1.38 billion with a $50 million margin of error, non-GAAP operating expenses of about $780 million, and non-GAAP fully diluted earnings per share of $0.47 with a $0.04 margin of error, with results due by November 4. The stock trades at around 110x forward earnings, and short interest stood at 11.52% of the public float as of August 31, while 52 hedge fund portfolios held a stake in the second quarter, up from 46 in the prior period. Cramer argued that panicking over steep pullbacks in dominant tech leaders is a losing game, calling short-term drops buying opportunities for patient investors given Arm's royalty growth and tight supply.

Bank of America estimates Meta Platforms could save roughly $8.5 billion in 2027 by running AI workloads on its own custom silicon instead of buying third-party chips, an outside analyst estimate rather than company guidance. The figure rests on a specific roadmap: Meta plans to deploy its third-generation MTIA 450 chip, code-named Arke, in the first half of 2027, followed by the higher-performance MTIA 500, or Astrid, later that year, both co-developed with Broadcom and aimed at AI inference workloads. BofA models Meta deploying 5 to 6 gigawatts of owned capacity in 2027 at a total cost of roughly $200 billion, assumes chips make up 60% of that spend, and pegs Meta's custom silicon as about 40% cheaper than third-party equivalents. Broadcom CEO Hock Tan said custom chips optimized for a customer's own workloads outperform any GPU and can do so at half the cost, and confirmed Broadcom will deliver three generations of MTIA accelerators to Meta between now and the end of 2027. Meta's FY2026 capex guidance sits at $130 billion to $145 billion, narrowed from $125 billion to $145 billion, with total expense guidance raised to $165 billion to $169 billion, while Q2 2026 revenue reached $60.80 billion, up 27.96% year over year, on advertising revenue of $59.36 billion.

Microchip Technology said its Silicon Storage Technology subsidiary has secured AnalogAI as a customer for its memBrain SAGE intellectual property, a design win that expands the company's exposure to edge AI and neuromorphic computing. The technology combines analog compute-in-memory with ultra-low-power operation to support real-time AI inference and learning directly on edge devices, and AnalogAI plans to use it in processors targeting humanoid robots, drones and vehicles. The win builds on Microchip's technology-licensing momentum, with licensing revenues rising to $42.6 million in the first quarter of fiscal 2027 from $33 million a year earlier, and licensee revenues growing to $163.8 million in fiscal 2026 from $131.1 million in fiscal 2025. Microchip faces intensifying competition in edge AI from NXP Semiconductors, which said AI-enabled processors are expected to represent 15% of industrial and IoT processor revenues in 2026, more than doubling from 2025, and from Texas Instruments, which cited strong demand across robotics, industrial automation and energy infrastructure. Microchip shares have risen 9.2% year to date, trailing the broader Zacks Computer and Technology sector's 16.7% gain, and the stock trades at a forward 12-month price/earnings of 17.31X versus the sector's 20.32X. The Zacks Consensus Estimate for Microchip's earnings stands at 92 cents per share, up a couple of cents over the past 30 days and implying 162.86% growth.

Arm CEO Rene Haas said demand is running at record strength across edge, automotive, robotics and data center, and that near-term revenue growth is capped by a multi-stage manufacturing bottleneck spanning wafers, substrates, testers and memory rather than by any softening in orders. Data-center royalty revenue more than doubled year over year last quarter, and the most recent 500 million Neoverse cores shipped in nine months after the first billion took six years, while IDC reported spending on Arm-based accelerated server platforms nearly doubled over the past two quarters and surpassed x86. Haas said his confidence in hitting the $2 billion AGI CPU target rose from May to July and again from July to September, after he earlier cut that target to $1 billion on foundry capacity worries, and Arm has secured the manufacturing capacity needed to support the $1 billion opportunity for the AGI CPU. The AGI CPU, Arm's first in-house data center chip, carries first-generation gross margin guidance in the high 30% range to low 40s with a path to 50% over the next couple of years, below core royalty margins, so silicon revenue will initially dilute reported profitability. Arm shares trade at $243.98, down 10.11% over the past month and 7.66% in the last week but up 123.2% year to date, at a trailing P/E near 247x, after Q1 FY2027 EPS of $0.25 missed the $0.40 estimate and operating margin compressed to 7% from 11%; analysts carry an average target of $288.70.

China has unveiled a new five-year development plan for manufacturing in electronics- and information technology-related industries, setting a target to raise revenue above 30 trillion yuan, or 4.5 trillion dollars, by 2030. The plan was drawn up jointly by the Ministry of Industry and Information Technology, or MIIT, and the National Development and Reform Commission, or NDRC. It covers the strategic framework from 2026 to 2030 and reflects the Chinese government's push for supply chain self-reliance, aiming for global leadership in strategically important emerging technologies from artificial intelligence to advanced computing, amid ongoing global competition. Beyond the revenue target, China also aims to raise research and development investment in the industry to 3.5% by 2030, setting out 17 missions covering nine key industry groups, from electronic components and materials, specialised manufacturing and measurement equipment, photonics, advanced computing, consumer electronics, power electronics, positioning and timing information technology, chips and AI endpoint devices, through to industrial electronics. The plan also drives technology development in the integrated circuit industry supply chain, while upgrading advanced electronic materials and smart sensors. In computing and memory, China will develop high-bandwidth flash memory alongside new forms of storage products such as Ferroelectric RAM, Resistive RAM and Magnetoresistive RAM, and aims to build very large-scale intelligent computing infrastructure supporting AI clusters equipped with hundreds of thousands of accelerator cards, while accelerating the deployment of space computing technology. The announcement of this sector-specific plan follows China's release on September 7 of its 15th five-year development plan for the information and communications industry, drawn up by the Ministry of Industry and Information Technology.

M31 Technology has expanded its technology collaboration with Rambus Inc. to cover MIPI CSI and MIPI DSI Controller technologies, combining them with M31's silicon-proven MIPI PHY portfolio to offer a comprehensive MIPI subsystem solution. The Hsinchu-based silicon IP provider said the integration of high-speed MIPI PHYs with MIPI CSI and DSI controllers lets customers reduce verification and development risks tied to multi-vendor integration while accelerating time-to-market. M31 CEO Scott Chang said customers previously had to procure controllers separately and then rely on M31 for integration, but can now entrust the complete MIPI interface design to M31. As M31 moves from PHY-only offerings to complete subsystem solutions that include controllers, the average total contract value of related engagements is expected to increase compared with the previous PHY-only licensing model. The company cited rising demand across smartphones, automotive ADAS cameras, multi-display cockpit systems, and AI edge devices, and said industry analysts project the global interface IP market will reach approximately US$5.4 billion by 2029.

Meta Platforms is planning to increase its use of in-house-designed AI chips as it looks to reduce the cost of its vast infrastructure buildout. The company has invested billions in data centers, servers and accelerators as it builds more sophisticated models and rolls out AI features across Facebook, Instagram and WhatsApp. Custom silicon can be tuned around particular workloads rather than the wide range of tasks expected from general-purpose accelerators, making AI services cheaper and more efficient to run at massive scale. More internal chips would lower Meta's reliance on outside semiconductor providers for some tasks, even as sophisticated Nvidia accelerators remain core to its infrastructure. The move follows a wider trend among hyperscalers such as Google and Amazon that have developed specialized processors for AI and cloud computing, and Meta's next test is whether those chips can deliver considerable savings as its AI infrastructure spending keeps growing.

Arm Holdings shares surged approximately 3.1% to $246.64 Tuesday morning as AI-linked semiconductor stocks rebounded from Monday's selloff. The company reported record fiscal first-quarter revenue of $1.29 billion, up 22% from a year earlier, with royalty revenue climbing 22% to $715 million and licensing revenue rising 23% to $574 million. Data-center royalties more than doubled as cloud providers continued deploying Arm-based processors. Royalties accounted for roughly 55.4% of quarterly revenue, though the stock trades 27.67% above its GF Value estimate of $193.19.

Marvell Technology's data center revenue reached $2.17 billion last quarter, up 46% year-over-year and now accounting for 79% of total company revenue, as the chipmaker positions itself as a key player in the AI infrastructure buildout. Management raised its fiscal 2027 revenue outlook to roughly $12 billion and expects fiscal 2028 data center revenue to grow more than 60%, with custom silicon revenue set to over-double next year. The company's expanded Google agreement validates its custom XPU thesis, and Marvell called scale-up optics a massive new TAM. Wall Street's consensus price target sits at $284.64 with 89% bullish coverage, while fiscal 2028 EPS estimates have climbed from $6.1726 ninety days ago to $6.7209 today. Shares trade at $218.07, up 156.98% year-to-date, though the stock fell 7.64% in a single session as part of a broad chip-sector reset. CEO Matt Murphy said AI-related bookings remain exceptionally robust and expects revenue growth to accelerate further through the remainder of fiscal 2027.

Cadence announced that its Tensilica IP Group and Analog Devices collaborated to develop the latest generation of SHARC automotive audio processors, bringing Cadence's Tensilica DSP architecture into ADI's next-generation ADSP product family. ADI is introducing the new ADSP-SC84x/2184x platform, which transitions from 28nm to 16nm process technology and builds on the ADSP-SC59x/2159x family. A key innovation of the ADSP-SC84x family is the new SHARC-FX processor core, designed and built through a joint engagement with Cadence, which delivers 5X higher performance compared to the previous generation SHARC+ processor used in the ADSP-SC59x family. The SHARC-FX features a custom DSP architecture supporting a broader range of data types, an enhanced VLIW SIMD architecture, efficient cache operations, and an expanded instruction set optimized for modern signal processing and AI workloads, including support for optimized neural-network kernel routines for frameworks such as Google's LiteRT. The platform also adds an LPDDR4 memory interface, increased on-chip L2 memory, enhanced cybersecurity via a dedicated hardware security module, and two optional connectivity cores with mainline Linux support.

Qualcomm has struck a long-term partnership with Amazon under which Amazon could purchase as much as $60 billion of Qualcomm's AI data-center chips and related products. As part of the agreement, Qualcomm granted Amazon warrants worth roughly $4 billion, allowing Amazon to buy Qualcomm shares at $161.26 per share, with the warrants vesting as Amazon purchases Qualcomm products. The two companies will develop custom AI chips focused on AI inference, along with high-speed optical connectivity for data centers, and Qualcomm is targeting $15 billion in annual data-center chip revenue by 2029. Amazon joins Microsoft and Meta as major cloud customers working with Qualcomm on custom AI chips, though the $60 billion figure is a potential purchase commitment rather than guaranteed revenue, and Qualcomm still must commercialize chips that can compete with Nvidia. Amazon's annualized custom-chip revenue run-rate exceeded $25 billion at the end of the June quarter, and Reuters reported separately that Amazon was preparing its first sterling bond sale as hyperscalers seek funding for AI infrastructure costs.

Piper Sandler launched coverage of the artificial intelligence chip sector on Thursday, initiating five semiconductor stocks at Overweight as prime beneficiaries of surging AI compute demand. Analyst David O'Connor started Nvidia with a $300 price target, calling it the "outright leader in AI compute with 80% market share" and describing the stock as "among the cheapest in AI universe" at about 14 times fiscal 2028 estimates. Broadcom was initiated at $460 as the leader in custom ASIC chips with a 75% share, where O'Connor wrote that "demand is 2x supply" and noted line of sight to 12 gigawatts of demand in fiscal 2027. Advanced Micro Devices was started at $600, an "Agentic AI Sweetspot" with earnings forecast to grow at a 65% compound annual rate through 2030, while Marvell Technology was initiated at $270, citing a $120 billion Google agreement and an Oct. 6 analyst day as a potential catalyst. Arm Holdings rounded out the group at $320, with O'Connor highlighting its CPU intellectual property dominance and expansion into custom silicon and accelerator IP that he said could double earnings.

According to SNS Insider, the global Gate-All-Around (GAA) transistor market is projected to grow from USD 685.00 million in 2025 to USD 2,283.00 million by 2035, at a CAGR of 12.8% from 2026 to 2035. Asia Pacific led the market with a 71.0% share in 2025, while North America is the fastest-growing region at approximately 16.5% CAGR, driven by CHIPS Act-backed fab investment. Within the U.S., the market is expected to rise from about USD 108.00 million in 2025 to USD 476.00 million by 2035. Nanosheet GAA transistors held the largest device-type share at 44%, with forksheet FET the fastest-growing type. Among technology nodes, 3nm/2nm nodes accounted for 40% of the market, while high-performance computing led applications with a 35% share, and foundries dominated end-users at 50%. Key players include TSMC, Samsung, Intel, and others, with TSMC ramping its N2 process and Intel advancing RibbonFET technology.

According to SNS Insider, the global semiconductor design market is projected to grow from USD 25.23 billion in 2025 to USD 69.43 billion by 2035, at a CAGR of 10.67% from 2026 to 2035. The market, which provides intellectual property, engineering expertise, and EDA solutions for logic, memory, and analog ICs, is forecast to reach USD 27.92 billion in 2026, driven by demand for AI accelerators and high-performance computing chips. Asia Pacific holds the largest regional share at approximately 39% in 2025 and is also the fastest-growing region, led by China, Taiwan, South Korea, Japan, and India. Within the U.S., the market is expected to grow from USD 7.89 billion in 2025 to USD 20.75 billion by 2035, a CAGR of about 10.12%, supported by hyperscaler investments and government initiatives like the CHIPS and Science Act. Key developments include NVIDIA's Blackwell Ultra GPU and Qualcomm's Snapdragon 8 Elite Gen 5, both launched in 2025, while Synopsys enhanced its AI-assisted EDA suite in 2024.

OpenAI CFO Sarah Friar confirmed at Goldman Sachs' Communacopia conference on September 8 that the company used its own frontier AI models to design the Jalapeno custom inference chip, marking the first confirmed instance of an AI lab using its own intelligence to architect its silicon. The chip, co-developed with Broadcom and announced in June 2026, achieved a nine-month design-to-tape-out cycle, the fastest ever reported for high-performance hardware, though the figure is company-claimed. This move is tied to OpenAI's aggressive pricing, including an 80% reduction for the Luna model to $0.20 per million input tokens and $1.20 per million output tokens, which has driven a 10x usage increase and supports 32% enterprise revenue growth from June to July 2026. By controlling the silicon stack, OpenAI aims to decouple margins from third-party cloud providers, reinforcing its position as a full-stack utility ahead of its confidential S-1 filing with an $852B post-money valuation. Broadcom CEO Hock Tan indicated small prototypes are expected in late 2026, with full-scale ramp slated for the first half of 2028, a period that will test the scalability of this recursive design model.

Shares of Applied Materials, AMD, Broadcom, Lam Research, and Marvell Technology surged in afternoon trading after Qualcomm announced a multi-generational collaboration with Amazon Web Services to develop custom AI data center chips, and ASML, TSMC, and Intel revealed key advancements in next-generation chip manufacturing. Qualcomm's partnership with AWS focuses on co-designing custom AI chips for high-performance inference and data center workloads, aiming to cut operating costs and improve energy efficiency. At the SPIE Photomask conference, ASML and TSMC launched an initiative to transition from 6-inch to 12-inch photomasks, targeting a pilot line by 2031 and full production by 2033, which would reduce costs and boost output for advanced AI processors. Additionally, Intel Foundry and ASML confirmed that Intel has processed over one million wafers on its Intel 18A node using High NA EUV equipment. These developments boosted investor confidence in sub-2-nanometer chip designs and AI infrastructure, with Applied Materials up 3.4%, AMD up 6.5%, Broadcom up 3.3%, Lam Research up 4.2%, and Marvell Technology up 3.3%.

Morgan Stanley has adopted a more selective stance on European semiconductor stocks, maintaining an overall positive outlook while adjusting ratings and price targets as the memory cycle nears a turn. Analyst Lee Simpson noted that strong AI demand and a broadening cycle recovery underpin the sector, but with DRAM approaching a late-cycle inflection and valuation dispersion widening, the bank is becoming more selective. The bank upgraded Synopsys to Overweight, moved Infineon to Equal Weight, and trimmed price targets on ASML and BE Semiconductor. ASML's target was lowered to €1,700 from €1,930, Infineon's to €65 from €81, and Besi's to €220 from €260. Simpson highlighted that memory is the key concern, with DRAM pricing near a peak and the cycle expected to turn by the fourth quarter. Despite a volatile year, with the sector up about 70% before retreating around 60% from June peaks, semis remain among the most-owned sectors, and the pullback may offer attractive entry points. Risks into year-end include potential shortfalls in data center buildout, a tougher macro backdrop, and delays in 800-volt vehicle architecture.

Sands Capital's Global Growth Fund added Arm Holdings to its portfolio in the second quarter of 2026, citing rising CPU demand from agentic AI as a key driver. The fund expects Arm's CPU royalties to grow from about 10 percent of revenue in 2026 to over 30 percent by 2031, leveraging its energy-efficient architecture to win cloud hyperscaler customers while maintaining over 90 percent share in smartphones. Arm shares closed at $252.09 on September 4, 2026, with a market capitalization of $269.23 billion, having gained 81.18 percent over the past year. The fund's portfolio returned 22.2 percent in the quarter, outperforming the MSCI All Country World Index's 14.9 percent, with gains driven by AI infrastructure earnings growth.

QuickLogic has unveiled its next-generation eFPGA Hard IP for GlobalFoundries 12LP, delivering improved performance and area efficiency while scaling beyond 250,000 LUTs. The enhanced IP adds a new DSP Multiply-and-Accumulate unit with enhanced preadder, cascade connections, and SIMD functionality, supporting applications such as phased array radar, electronic warfare, wireless infrastructure, satellite communications, and imaging algorithms. It also introduces an optional Configuration Bit Health Monitor that detects and corrects configuration bit errors induced by Single Event Upsets during active operation, enhancing reliability for demanding applications. QuickLogic's eFPGA Hard IP allows programmable logic to be embedded directly within an SoC, enabling post-manufacturing updates to hardware functions, and is supported by the company's Aurora and AuroraPRO tools. QuickLogic will showcase the new IP as a Silver Sponsor at the GlobalFoundries Technology Summit North America on September 2, 2026, in Santa Clara, California.

Cadence announced that its PHY and controller IP for the PCIe 6.0 specification, implemented in TSMC's N3 process, achieved first-pass success at the PCI-SIG compliance workshop in late July, passing all official tests at 64 GT/s and earning a place on PCI-SIG's Integrators List. The complete x8 subsystem solution, which includes both PHY and controller, is now available for SoC providers to design in, demonstrating industry readiness for mass production and de-risking adoption for AI/HPC customers. Cadence's vice president of R&D for high-performance PHY, Marc Loinaz, highlighted the solution's best-in-class performance and low power, while Keysight's Brig Asay and PCI-SIG's Al Yanes praised the milestone. Positron AI has licensed the SerDes IP for its AI inference accelerator chip, citing confidence in the subsystem's silicon-proven design and interoperability testing. The PCIe 6.0 standard addresses growing demand for high-bandwidth connectivity in AI and high-performance computing systems, with expected wide adoption in data centers and AI infrastructure.

Primarius Technologies disclosed on the evening of August 31 that its plan to acquire 100% equity in Analog Semiconductor and 45.64% equity in Nano Microelectronics through share issuance and cash payment, along with raising supporting funds, has made new progress. The issue price for this supporting fundraising is set at 38.92 yuan per share, with 26.9784 million shares to be issued, raising a total of 1.05 billion yuan. Two investors have been confirmed, with Guotou Jixin receiving 24.409 million shares, accounting for 90.5% of the total, and an allocation amount of 950 million yuan. According to Qichacha, Guotou Jixin has a registered capital of 71.071 billion yuan, with Big Fund Phase III holding a 99.9% stake. As the first listed EDA company in China, Primarius Technologies aims through this acquisition to integrate the semiconductor IP resources of Analog Semiconductor and Nano Microelectronics, creating the country's first full-link empowerment platform with dual-engine power in EDA and IP. According to its semi-annual report, the company's operating revenue in the first half of the year increased by 10.05% year-on-year, but net profit attributable to the parent company was a loss of 6.7313 million yuan, mainly due to changes in fair value and reduced investment income.

Cadence Design Systems raised its 2026 revenue guidance to $6.26-$6.34 billion from $6.125-$6.225 billion, citing broad-based demand for its AI-driven portfolio. In the second quarter of 2026, the company generated $635 million in operating cash flow and $582 million in free cash flow, up from $356 million and $307 million in the prior quarter, respectively. Cadence repurchased $200 million of shares in the quarter and expects to use roughly 50% of its free cash flow for buybacks in 2026. The company also raised its 2026 operating cash flow forecast to $2 billion from $1.875-$1.975 billion. Management expects acquisition and IP profitability to improve into 2027, and in February 2026, Cadence acquired Hexagon AB's Design & Engineering division for €2.7 billion to expand in the structural analysis market. Despite concerns over macroeconomic volatility and competition from Synopsys and Keysight, Cadence remains asset-light and does not expect Agentic AI to significantly increase capital intensity.

Marvell Technology reported fiscal second-quarter data center revenue of $2.1715 billion, up 46% year over year, and total revenue of $2.739 billion, up 36.55%, with management guiding to roughly 50% growth in fiscal 2028. The company also disclosed an expanded custom silicon agreement with Google that includes a warrant allowing Google to acquire up to 7% of Marvell's shares tied to revenue milestones, with one analyst estimating the deal could generate $120 billion in cumulative revenue over six years. Non-GAAP operating margin expanded to 36.6% from 34.8%, cash surged 221.2% to $3.93 billion, and the company repurchased $200 million of stock in the quarter. Marvell's forward P/E stands at 60, and shares are down 8.61% over the past week but up 155.27% year to date at $216.62, with an Investor Day scheduled for October 6, 2026.

Synopsys reported fiscal third-quarter revenue of $2.477 billion, non-GAAP operating margin of 41.6%, and non-GAAP EPS of $3.91, all exceeding the high end of guidance, driven by strength in EDA and Ansys. The company raised its full-year revenue guidance to $9.69-$9.74 billion, with non-GAAP EPS of $15.04-$15.10, and now expects double-digit organic EDA growth for the fourth quarter and full year. Backlog stands at $10.9 billion, down modestly due to the divestiture of the processor IP business. CEO Sassine Ghazi highlighted the launch of Multiphysics Fusion, the first joint Synopsys-Ansys solution, and noted strong demand for agentic AI and custom silicon opportunities. CFO Shelagh Glaser cited record hardware revenue and ahead-of-schedule Ansys cost synergies, with free cash flow guidance raised to approximately $2.6 billion.

IBM has introduced its next-generation dual-architecture processor for IBM Z and LinuxONE systems, marking the first hardware milestone from its partnership with Arm Holdings announced in April. The chip, built on a 2-nanometer process, features 11 high-performance cores running above 5.7 GHz, AI inference accelerators for fraud detection, and enhanced security features. This collaboration aims to expand IBM's mainframe software ecosystem by supporting Arm workloads, while giving Arm a foothold in enterprise computing. Both companies have seen increased hedge fund interest, with IBM's hedge fund holders rising from 59 to 74 and Arm's from 46 to 52 in the second quarter. The success of this venture hinges on customer adoption of Arm workloads on mainframes, a key point to watch in upcoming earnings calls.

The implementation of the AI4Chip policy, combined with Nvidia's earnings beating expectations and confirming computing power demand, drove the Science and Technology Innovation Chip Index tracked by Southern Fund's Science and Technology Innovation Chip ETF Southern, ticker 588890, to close up 4.55% yesterday. As of the close on August 27, 2026, the ETF's turnover rate was 4.09%, with a trading volume of 152 million yuan. On the policy front, the Beijing Economic-Technological Development Area issued the nation's first AI4Chip special action plan, using five major actions and 20 tasks to promote AI empowerment in chip design, EDA tools, and manufacturing. On the same day, Shanghai issued its 15th Five-Year Plan for new industrialization, clearly targeting the entire chain in integrated circuits. On the industry front, Nvidia's second-quarter revenue reached 96.2 billion dollars, up 106% year on year, with data center revenue of 89 billion dollars, up 117% year on year. Its third-quarter guidance of 108 billion dollars exceeded expectations, confirming global AI computing power demand. In addition, price increase signals in semiconductor materials are clear, with silicon wafer prices up 18% to 22%, tungsten hexafluoride exceeding 3.15 million yuan per ton, and target materials up as much as 70%. The industry is currently at a critical juncture as AI computing power demand spreads from the training side to the inference side, with domestic AI accelerator card shipment share already rising to 41%, and Goldman Sachs predicting it could exceed 50% in 2026.