Edge AI Silicon Ignition: ARM Secures .85B Q2 Licensing Surge, NPU Market Poised for 35% CAGR to $42B by 2028; NVDA & TSM Lead Next-Gen Mobile AI Supercycle
Arm Holdings (ARM) reports a +28.5% year-over-year Q2 2026 revenue jump to .92B, fueled by robust NPU and data center licensing. TSMC (TSM) 3nm utilization holds at 95%, while HBM3e contract prices surge +20-25% driving SK Hynix (000660.KS) operating margins up 350bps.
Tradesnaut Quant Research Desk · September 17, 2026 · 6 min read · Semiconductors
.85B Q2 Licensing Surge, NPU Market Poised for 35% CAGR to $42B by 2028; NVDA & TSM Lead Next-Gen Mobile AI Supercycle" />
Key takeaways
- ARM Holdings (ARM) Q2 2026 revenue reached .92B, a +28.5% YoY increase, with NPU IP licensing driving a significant portion of the .85B licensing segment, indicating strong demand for efficient AI compute at the edge.
- The global Edge AI silicon market is on track to expand at a 35% CAGR, reaching $42B by 2028, largely propelled by a mobile device upgrade supercycle and surging demand for dedicated Neural Processing Units (NPUs) in client devices and compact data centers.
- HBM3e contract prices have escalated by +20-25% in Q3 2026, boosting memory manufacturers like SK Hynix (000660.KS) and Samsung Electronics (005930.KS), with SK Hynix reporting a 350 basis point improvement in its operating profit margin for memory products in the same period.
Market Dynamics & Earnings Data Breakdown
The semiconductor industry is entering a profound device upgrade supercycle, primarily driven by the integration of advanced Neural Processing Units (NPUs) across mobile, edge, and compact data center segments. Arm Holdings (ARM) reported stellar Q2 2026 results, with total revenue climbing +28.5% year-over-year to
.92 billion, comfortably beating analyst consensus estimates of .80 billion. The core of this growth stems from its licensing division, which saw revenue jump +32% to .85 billion, largely attributed to surging demand for its high-performance Cortex-X and Cortex-A series CPU IP integrated with its Mali-G and Ethos NPU architectures for next-generation edge AI platforms. This robust performance underscores Arm's pivotal role as the foundational IP provider for an estimated 90% of the burgeoning edge AI silicon market, projected to reach 5 billion in 2026 and accelerate to $42 billion by 2028 with a formidable 35% compound annual growth rate (CAGR).
Nvidia (NVDA) continues to demonstrate its market leadership, with NPU-related revenue from its Jetson and inference platforms contributing an estimated $5.5 billion in H1 2026, representing a +45% increase from the prior year. While server-side AI accelerators remain its profit engine, Nvidia’s strategic push into the edge with its AI-enabled System-on-Chips (SoCs) for IoT, robotics, and industrial applications is yielding significant returns, bolstering its enterprise segment's operating profit margin by 250 basis points to 62.5%. Meanwhile, Qualcomm (QCOM) reported that its Snapdragon 8 Gen 4 and Gen 5 mobile platforms, featuring enhanced NPUs, powered over 180 million premium smartphones shipped in H1 2026, reflecting a +18% increase in NPU-enabled device penetration. This momentum is further validated by independent market data showing that devices with dedicated NPUs are commanding an average selling price (ASP) premium of 12-15% over their non-NPU counterparts, reinforcing the thesis of a value-driven upgrade cycle.
While Arm's ecosystem dominates, early indicators suggest increasing traction for RISC-V architectures in specific edge AI niches. Companies like Google Cloud (GOOGL) and Amazon AWS (AMZN) are exploring RISC-V for custom inference accelerators in their edge infrastructure, aiming for lower power consumption and greater customization. Although RISC-V's overall market share remains below 5% for NPU-enabled devices, its adoption in specialized embedded systems and IoT devices grew +50% in H1 2026, albeit from a smaller base. The industry is closely watching developments in this space, as RISC-V could introduce cost efficiencies of up to 20-30% in certain low-power, high-volume edge deployments, potentially challenging Arm’s long-term dominance in specific segments after 2028.
Supply Chain Bottlenecks & Macro Valuation Metrics
The escalating demand for high-performance memory and advanced packaging required by next-generation NPUs is creating noticeable bottlenecks across the supply chain. HBM3e (High Bandwidth Memory 3e) contract prices have surged an unprecedented +20-25% for Q3 2026, directly benefiting memory giants like SK Hynix (000660.KS) and Samsung Electronics (005930.KS). SK Hynix, a primary supplier of HBM3e for Nvidia’s AI platforms, reported a 350 basis point expansion in its overall memory operating profit margin, reaching 28.1% for Q2 2026, driven by this pricing power and robust demand for its LPDDR5X mobile DRAM with integrated NPU support. Micron Technology (MU) also saw its LPDDR5X revenues grow +15% QoQ, with gross margins for its high-density mobile solutions improving by 400 basis points.
Foundry capacity remains critically tight for cutting-edge nodes. TSMC (TSM) reported 95% utilization rates for its 3nm process technology in Q2 2026, primarily driven by orders for Apple, Nvidia, and Qualcomm’s NPU-accelerated SoCs. Capital expenditure (capex) across the semiconductor industry is projected to exceed 50 billion in 2026, with TSMC alone investing $40-42 billion to expand its advanced node capacity. This massive investment underscores the industry’s commitment to meeting the insatiable demand for AI silicon, yet lead times for certain specialty equipment from ASML (ASML), particularly EUV tools, extend beyond 18 months, indicating persistent supply-side constraints into late 2027. Despite these capex figures, analysts project that the industry's aggregate capacity additions may only meet approximately 85% of the forecasted NPU demand by mid-2027.
Valuation metrics for key players reflect this bullish outlook. Nvidia (NVDA) is trading at a forward P/E of 35x 2027 estimated earnings and an EV/EBITDA of 28x, still commanding a significant premium given its AI leadership and pipeline. Arm Holdings (ARM), with its strategic IP advantage, trades at a lofty 50x EV/EBITDA multiple for 2027, reflecting its licensing model's high leverage and scalability. By contrast, more diversified players like Broadcom (AVGO), which also contributes to edge AI with custom silicon and connectivity, are trading at a more modest 22x forward P/E, suggesting a potential undervalued opportunity for investors seeking broader semiconductor exposure. Institutional capital flows into semiconductor ETFs (e.g., SOXX) have exceeded
4.2 billion year-to-date, indicating strong conviction in the sector’s long-term growth trajectory.
Quantitative Order Flow & Volatility Metrics
Quantitative analysis of options order flow reveals a strong bullish bias for NPU-centric semiconductor names. For Nvidia (NVDA), the 3-month at-the-money (ATM) call skew is currently at +2.5%, indicating higher demand for upside calls relative to puts, a signal of strong speculative interest in further price appreciation. Weekly options volume for NVDA has averaged over 1.8 million contracts in September 2026, with a significant 65% tilt towards calls, reinforcing this positive sentiment. Similarly, the SOX Semiconductor Index (SOX), a key barometer for the sector, has surged +28% year-to-date, hitting a new all-time high of 6,250 points, with institutional net buying contributing over $8.7 billion in the last quarter alone, predominantly targeting advanced chipmakers and memory providers.
The implied volatility for the SMH ETF (VanEck Semiconductor ETF) has risen by +15% over the past month, reaching 28.5%, reflecting increased uncertainty but also strong price momentum around upcoming product announcements and earnings. This elevated volatility, however, is being met with robust institutional buying pressure, particularly in longer-dated call options for Arm Holdings (ARM) and TSMC (TSM). For ARM, open interest in January 2027 calls with strike prices above
70 has increased by 120,000 contracts over the past three weeks, highlighting conviction in its sustained growth trajectory.
Memory chip manufacturers, while more cyclical, are also showing signs of positive order flow. SK Hynix (000660.KS) saw its daily call options volume on the Korea Exchange increase by +120% last week following the HBM3e contract price reports, with a call/put ratio of 1.8x, suggesting that options traders are pricing in continued upside for memory-related revenues. The KOSPI index, heavily influenced by Samsung Electronics and SK Hynix, has gained +7.2% year-to-date, reflecting the broader positive sentiment towards the region's semiconductor leaders. The overall market risk premium, as measured by the VIX index, remains suppressed at 13.5, indicating a generally constructive environment for risk assets, although specific sector volatility is elevated due to rapid technological shifts.
Tags: Memory Chips, SK Hynix, Semiconductors, Wall Street, Edge AI, NPU, ARM Holdings, Nvidia, TSMC, RISC-V
Arm Holdings (ARM) reports a +28.5% year-over-year Q2 2026 revenue jump to Tradesnaut Quant Research Desk · September 17, 2026 · 6 min read · Semiconductors The semiconductor industry is entering a profound device upgrade supercycle, primarily driven by the integration of advanced Neural Processing Units (NPUs) across mobile, edge, and compact data center segments. Arm Holdings (ARM) reported stellar Q2 2026 results, with total revenue climbing +28.5% year-over-year to Nvidia (NVDA) continues to demonstrate its market leadership, with NPU-related revenue from its Jetson and inference platforms contributing an estimated $5.5 billion in H1 2026, representing a +45% increase from the prior year. While server-side AI accelerators remain its profit engine, Nvidia’s strategic push into the edge with its AI-enabled System-on-Chips (SoCs) for IoT, robotics, and industrial applications is yielding significant returns, bolstering its enterprise segment's operating profit margin by 250 basis points to 62.5%. Meanwhile, Qualcomm (QCOM) reported that its Snapdragon 8 Gen 4 and Gen 5 mobile platforms, featuring enhanced NPUs, powered over 180 million premium smartphones shipped in H1 2026, reflecting a +18% increase in NPU-enabled device penetration. This momentum is further validated by independent market data showing that devices with dedicated NPUs are commanding an average selling price (ASP) premium of 12-15% over their non-NPU counterparts, reinforcing the thesis of a value-driven upgrade cycle. While Arm's ecosystem dominates, early indicators suggest increasing traction for RISC-V architectures in specific edge AI niches. Companies like Google Cloud (GOOGL) and Amazon AWS (AMZN) are exploring RISC-V for custom inference accelerators in their edge infrastructure, aiming for lower power consumption and greater customization. Although RISC-V's overall market share remains below 5% for NPU-enabled devices, its adoption in specialized embedded systems and IoT devices grew +50% in H1 2026, albeit from a smaller base. The industry is closely watching developments in this space, as RISC-V could introduce cost efficiencies of up to 20-30% in certain low-power, high-volume edge deployments, potentially challenging Arm’s long-term dominance in specific segments after 2028. The escalating demand for high-performance memory and advanced packaging required by next-generation NPUs is creating noticeable bottlenecks across the supply chain. HBM3e (High Bandwidth Memory 3e) contract prices have surged an unprecedented +20-25% for Q3 2026, directly benefiting memory giants like SK Hynix (000660.KS) and Samsung Electronics (005930.KS). SK Hynix, a primary supplier of HBM3e for Nvidia’s AI platforms, reported a 350 basis point expansion in its overall memory operating profit margin, reaching 28.1% for Q2 2026, driven by this pricing power and robust demand for its LPDDR5X mobile DRAM with integrated NPU support. Micron Technology (MU) also saw its LPDDR5X revenues grow +15% QoQ, with gross margins for its high-density mobile solutions improving by 400 basis points. Foundry capacity remains critically tight for cutting-edge nodes. TSMC (TSM) reported 95% utilization rates for its 3nm process technology in Q2 2026, primarily driven by orders for Apple, Nvidia, and Qualcomm’s NPU-accelerated SoCs. Capital expenditure (capex) across the semiconductor industry is projected to exceed 50 billion in 2026, with TSMC alone investing $40-42 billion to expand its advanced node capacity. This massive investment underscores the industry’s commitment to meeting the insatiable demand for AI silicon, yet lead times for certain specialty equipment from ASML (ASML), particularly EUV tools, extend beyond 18 months, indicating persistent supply-side constraints into late 2027. Despite these capex figures, analysts project that the industry's aggregate capacity additions may only meet approximately 85% of the forecasted NPU demand by mid-2027. Valuation metrics for key players reflect this bullish outlook. Nvidia (NVDA) is trading at a forward P/E of 35x 2027 estimated earnings and an EV/EBITDA of 28x, still commanding a significant premium given its AI leadership and pipeline. Arm Holdings (ARM), with its strategic IP advantage, trades at a lofty 50x EV/EBITDA multiple for 2027, reflecting its licensing model's high leverage and scalability. By contrast, more diversified players like Broadcom (AVGO), which also contributes to edge AI with custom silicon and connectivity, are trading at a more modest 22x forward P/E, suggesting a potential undervalued opportunity for investors seeking broader semiconductor exposure. Institutional capital flows into semiconductor ETFs (e.g., SOXX) have exceeded Quantitative analysis of options order flow reveals a strong bullish bias for NPU-centric semiconductor names. For Nvidia (NVDA), the 3-month at-the-money (ATM) call skew is currently at +2.5%, indicating higher demand for upside calls relative to puts, a signal of strong speculative interest in further price appreciation. Weekly options volume for NVDA has averaged over 1.8 million contracts in September 2026, with a significant 65% tilt towards calls, reinforcing this positive sentiment. Similarly, the SOX Semiconductor Index (SOX), a key barometer for the sector, has surged +28% year-to-date, hitting a new all-time high of 6,250 points, with institutional net buying contributing over $8.7 billion in the last quarter alone, predominantly targeting advanced chipmakers and memory providers. The implied volatility for the SMH ETF (VanEck Semiconductor ETF) has risen by +15% over the past month, reaching 28.5%, reflecting increased uncertainty but also strong price momentum around upcoming product announcements and earnings. This elevated volatility, however, is being met with robust institutional buying pressure, particularly in longer-dated call options for Arm Holdings (ARM) and TSMC (TSM). For ARM, open interest in January 2027 calls with strike prices above Memory chip manufacturers, while more cyclical, are also showing signs of positive order flow. SK Hynix (000660.KS) saw its daily call options volume on the Korea Exchange increase by +120% last week following the HBM3e contract price reports, with a call/put ratio of 1.8x, suggesting that options traders are pricing in continued upside for memory-related revenues. The KOSPI index, heavily influenced by Samsung Electronics and SK Hynix, has gained +7.2% year-to-date, reflecting the broader positive sentiment towards the region's semiconductor leaders. The overall market risk premium, as measured by the VIX index, remains suppressed at 13.5, indicating a generally constructive environment for risk assets, although specific sector volatility is elevated due to rapid technological shifts. Tags: Memory Chips, SK Hynix, Semiconductors, Wall Street, Edge AI, NPU, ARM Holdings, Nvidia, TSMC, RISC-V.85B Q2 Licensing Surge, NPU Market Poised for 35% CAGR to $42B by 2028; NVDA & TSM Lead Next-Gen Mobile AI Supercycle" />
Key takeaways
Market Dynamics & Earnings Data Breakdown
Supply Chain Bottlenecks & Macro Valuation Metrics
Quantitative Order Flow & Volatility Metrics