Pavona’s Open‑Source Silicon Initiative Takes Shape

GlobalPlatform announced today that Pavona is now live as an open‑source silicon ecosystem. The consortium’s founding members—Meta, Qualcomm, Tenstorrent, Winbond, the University of Oxford and a host of analog, security and AI specialists—have pledged engineering resources and IP contributions to create a modular library of tape‑out‑proven, certification‑ready blocks. The goal, as the press release states, is to give hardware developers a reusable, vetted foundation comparable to what Linux provides for software.

---

Technical Foundations

Process‑node coverage

• 5 nm (FinFET) and 3 nm (GAA) blocks are already available, sourced from partners that have qualified designs on Samsung and TSMC fabs. Early benchmarks show a 15 % lower power envelope for a 64‑bit RISC‑V core built with Pavona’s 5 nm block versus a comparable proprietary block from 2022.
• 28 nm and 22 nm legacy nodes are included for IoT and automotive customers that cannot justify sub‑10 nm fabs. These designs retain the same security hardening (e.g., side‑channel mitigations) as the advanced nodes.

Core IP categories

Category	Example Block	Key Metrics	Certification Path
CPU	RISC‑V “P‑Core”	2.8 GHz @ 5 nm, 1.2 TOPS/W	Common Criteria EAL5, FIPS 140‑2
Accelerator	Tensor‑Flow Lite NN Engine	120 TOPS @ 3 nm, 0.8 W per 1 TOPS	ISO/IEC 62443
Security	Secure‑Boot ROM, Crypto‑Engine (AES‑256, SHA‑3)	<10 µs latency for 256‑bit key ops	Common Criteria EAL4+, EMVCo
Memory	2 MiB SRAM macro	0.9 pJ/bit read, 1.2 pJ/bit write	JEDEC JESD79‑4
Analog	Low‑noise ADC (12‑bit, 200 MS/s)	ENOB 11.2 bits, 2.5 mW	IEC 60747

All blocks are tape‑out proven: each has passed at least one silicon run in a production‑grade fab, with silicon‑validation data publicly available on the Pavona GitHub repository. The repository also hosts reference designs that stitch together a CPU, accelerator, and security subsystem into a single‑chip solution, complete with board‑level schematics and a Bill‑of‑Materials (BOM) that targets a $12 cm² die cost at 5 nm.

Certification‑ready workflow

1. Design import – IP is delivered as RTL (SystemVerilog) plus a pre‑validated physical‑design constraint set (PDK‑specific). The Pavona CI pipeline runs lint, formal verification and power‑analysis automatically.
2. Security hardening – A set of mandatory checks (e.g., constant‑time implementations, glitch‑filter insertion) is enforced before a block can be marked cert‑ready.
3. Documentation bundle – Each block ships with a security assessment report, test‑vector suite and a compliance checklist aligned with Common Criteria, FIPS and ISO standards.
4. Tape‑out package – For customers who lack internal fab access, Pavona partners with foundry‑as‑a‑service providers to deliver a turn‑key GDSII set and a fab‑ready mask set within 30 days of order.

The open‑source license is a dual‑licensing model: Apache‑2.0 for unrestricted use, and a commercial support add‑on that includes warranty, priority bug fixes and optional custom‑node porting.

---

Market Implications

Accelerating time‑to‑market

The average silicon startup now spends 18 months on IP integration before a first silicon tape‑out. Pavona’s pre‑qualified blocks cut that window to 6–9 months, a reduction of roughly 50 %. For a typical AI accelerator project, this translates to a $4 M reduction in NRE costs, based on the industry average of $8 M for a 5 nm design cycle.

Supply‑chain resilience

By providing multiple node options and a catalog of foundry‑agnostic designs, Pavona mitigates the current bottleneck that has forced many OEMs to queue for 5 nm capacity. Companies can fall back to 7 nm or 22 nm blocks without redesigning the security architecture, preserving product schedules while the high‑volume fabs clear backlog.

Competitive pressure on proprietary IP vendors

Major IP vendors such as Arm and Synopsys have historically priced high‑performance cores at $200 k–$300 k per license. Pavona’s open model, combined with optional commercial support at $30 k–$50 k per block, forces a price compression of 80 % for comparable functionality. Early adopters like Tenstorrent have already announced a next‑gen inference chip that will ship with a Pavona‑based security subsystem, citing a 30 % lower BOM versus their prior in‑house design.

Ecosystem growth prospects

The founding consortium includes 12 founding members and over 30 additional sign‑ups as of the announcement date. If the current onboarding rate holds—approximately 2 new members per month—the ecosystem could reach 50 contributors within a year, delivering an estimated 200 IP blocks and 50 reference SoCs. Such scale would create network effects similar to the Linux kernel’s contributor base, encouraging third‑party tool vendors to add Pavona‑aware synthesis and verification plugins.

---

Next Steps for Designers

1. Visit the portal – All IP, documentation and CI pipelines are hosted at Pavona.org.
2. Clone the repo – The main repository lives at github.com/pavona-ecosystem/ip-library and includes Docker‑based build scripts.
3. Run the compliance suite – A step‑by‑step guide shows how to generate a Common Criteria evidence package in under an hour.
4. Engage with the community – Monthly webinars, hosted by GlobalPlatform, will cover node‑porting strategies and security best practices.

Pavona’s launch marks a concrete step toward democratizing silicon design, offering a ready‑to‑fabricate, security‑first stack that can be adopted across everything from edge AI chips to automotive controllers. As the ecosystem matures, the balance of power may shift from a few proprietary IP holders to a broader, community‑driven model that mirrors the open‑source software revolution.