Beyond the FPGA-ASIC compromise: Runtime reconfigurable processors and the future of computing

How Morphing Machines is solving from India for the world a fundamental architectural bottleneck where software innovation is dramatically outpacing hardware development cycles

Modern computing faces an awkward paradox. Software evolves in months, while hardware takes years. A familiar example is that even an expensive smartphone purchased three years ago cannot unlock the full capabilities of today's applications. This isn’t because the silicon has degraded, but because of the combination of planned obsolescence in this industry and the rapid rate of evolution of software.

The best automakers in the world grapple with similar constraints, unable to deploy their latest autonomous driving algorithms across older vehicle fleets without costly hardware replacements. The fundamental issue is that chip architectures, once manufactured, remain frozen in time.

This rigidity stems from the traditional design trade-off between flexibility and performance. Field-programmable gate arrays, or FPGAs, offer adaptability — engineers can reconfigure them after manufacturing to suit new tasks. Yet this versatility comes at a cost: lower performance and higher power consumption compared to application-specific integrated circuits (ASICs). ASICs, by contrast, deliver peak efficiency for their designated workloads but cannot adapt once fabricated. For decades, chip designers have navigated this compromise, choosing one path at the expense of the other.

The stakes have risen considerably. Artificial intelligence workloads shift unpredictably; what worked for training large language models yesterday may prove suboptimal for inference-heavy applications tomorrow. Telecommunications infrastructure must accommodate evolving 5G and 6G standards. Autonomous vehicles require real-time processing that adapts to new sensor configurations and safety protocols. These demands expose the limitations of static hardware architectures.

Morphing Machines, a Bengaluru-based semiconductor startup, is working to resolve this tension. The company's REDEFINE processor architecture aims to combine FPGA-like reconfigurability with ASIC-level performance through runtime adaptability. Rather than requiring hardware replacement or lengthy reprogramming, the system dynamically allocates computational resources — shifting between CPU and GPU capacity cores depending on workload demands.

The technology builds on two decades of research at the Indian Institute of Science, where Professor SK Nandy developed foundational work in reconfigurable computing and system-on-chip design. His four decades of academic research, spanning application-specific instruction sets and high-performance architectures, established the scientific underpinning. Dr. Ranjani Narayan, who holds a PhD in computer architecture from IISc, contributed technical depth in parallel processing systems. In 2021, Deepak Shapeti joined as chief executive, sharpening their focus on commercial viability beyond technical elegance.

The company has demonstrated early capabilities on customer workloads and is now scaling toward proof-of-silicon. Morphing Machines raised $2.76 million in seed funding in June 2024, that we led, followed by a $4.3 million Series A round in October 2025 led by IAN Alpha Fund. The startup benefits from India's Design Linked Incentive scheme, part of the government's $10 billion semiconductor mission aimed at building domestic chip design capabilities.

Market conditions appear conducive, with the global FPGA market projected to grow from close to $12 billion in 2025 to about $19.3 billion by 2030, driven by demand for adaptable computing in AI acceleration, edge computing, and telecommunications infrastructure. Meanwhile, data centre chip spending continues its upward trajectory, with specialized processors gaining share as hyperscalers seek alternatives to general-purpose GPUs for cost-sensitive AI inference tasks.

Precedents exist for this approach. Cornami, a US-based reconfigurable computing venture, has raised over $120 million from investors including SoftBank and Applied Materials. SambaNova Systems, pursuing full-stack hardware-software integration for AI workloads, has attracted more than $1 billion in funding from SoftBank, Google, and Intel. These ventures signal investor appetite for architectures that bridge the flexibility-performance divide.

India's role in this shift merits attention. Industry experts reckon we are home to roughly one-fifth of the world's semiconductor design talent, yet has historically focused on services rather than intellectual property creation. Morphing Machines represents a test case: whether deep academic research, combined with government support and patient capital, can yield globally competitive processor designs. Success would not merely address a technical problem; it would demonstrate India's capacity to move from the periphery of semiconductor innovation to its centre.

The challenge remains formidable. Established players dominate the market, and chip design demands years of iteration before commercial deployment. But the fundamental tension — software racing ahead while hardware lags — is real, and growing more acute. Runtime reconfigurable processors may offer a path forward, provided the engineering can match the ambition.