Free tools Windows power users keep installed
One-click scans. No signup required.
Tenstorrent raised more than $693 million in a Series D round announced on December 2, 2024. The financing was led by Samsung Securities and AFW Partners, with Bezos Expeditions and other institutional investors participating. Headlines rounded the amount to $700 million and described the deal as Jeff Bezos and Samsung investing in an Nvidia challenger—but both descriptions need qualification.
The round gives Tenstorrent substantial capital to expand chip engineering, its supply chain, software ecosystem, and AI-server demonstrations. It does not mean Jeff Bezos personally invested a disclosed amount, that Samsung Electronics funded the entire round, or that Tenstorrent has already reached Nvidia’s scale, software maturity, revenue, or market share.
What happened in the Tenstorrent funding round?
Tenstorrent announced that it had closed more than $693 million in Series D financing on December 2, 2024. The company said the round was completed at a $2 billion pre-money valuation. A post-money figure sometimes reported as roughly $2.6 billion is an approximate calculation, not the valuation figure Tenstorrent officially disclosed.
The round was led by Samsung Securities and AFW Partners. Named participants included:
#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
- Bezos Expeditions
- XTX Markets
- Corner Capital
- Protagonist
- MESH
- Export Development Canada
- Healthcare of Ontario Pension Plan
- LG Technology Ventures
- Hyundai Motor Group
- Fidelity Management & Research Company
- Innovation Engine
- Baillie Gifford
Tenstorrent’s announcement says Barclays acted as the sole placement agent.
What the headline gets right—and wrong
The phrase “Jeff Bezos invests” is shorthand for the participation of Bezos Expeditions, Bezos’s investment firm. The announcement does not identify an individual contribution from Bezos or disclose how much Bezos Expeditions invested.
Likewise, “Samsung invests” is imprecise. The named lead investor was Samsung Securities. That should not automatically be read as a direct investment by Samsung Electronics or as proof of a manufacturing partnership with Samsung’s semiconductor division.
Finally, “$700 million” is a rounded headline figure. The precise company description is more than $693 million. These distinctions matter because a financing headline can create a stronger impression of corporate backing, personal involvement, or valuation certainty than the public filing actually supports.
What is Tenstorrent?
Tenstorrent is an AI-computing company led by CEO Jim Keller, a semiconductor engineer associated with major CPU and system-on-chip projects. It develops AI accelerators, RISC-V CPU technology, software, compiler tools, developer cards, and complete systems.
Its strategy is not simply to sell another standalone accelerator. Tenstorrent is attempting to offer an alternative computing stack that includes:
Rank #2
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
- Tensix processors for AI workloads.
- RISC-V CPU cores integrated into its broader platform.
- Ethernet-based networking for connecting accelerators and scaling systems.
- An open-source-oriented software and compiler ecosystem.
- Hardware aimed at both AI training and inference.
The company’s pitch is that customers should have more choice and visibility than they get from a CUDA-centered platform. But “open-source stack” does not mean that every hardware design, firmware component, commercial support service, or manufacturing process is open source.
Why Nvidia is the target
Nvidia’s advantage in AI computing extends well beyond the performance of an individual GPU. Its installed base, CUDA programming model, libraries, framework integrations, cloud availability, server partnerships, customer references, and supply-chain scale make it difficult for a new accelerator company to compete.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsFor many organizations, the main switching cost is software. A team may already depend on CUDA-specific kernels, optimized libraries, deployment tools, model-serving integrations, and engineers who understand Nvidia’s ecosystem. Even if another chip is less expensive, moving a production workload can require substantial porting, testing, debugging, and optimization.
Tenstorrent is therefore competing against an ecosystem, not only against Nvidia silicon. Its potential advantage is greatest where a buyer values hardware access, architectural flexibility, alternative software, or lower system costs and is willing to do more engineering work.
How Tenstorrent differs from Nvidia
| Issue | Nvidia’s position | Tenstorrent’s proposed alternative |
|---|---|---|
| Software | Mature CUDA ecosystem with extensive libraries and framework support | Open-source-oriented compiler and software components, with potentially more customer involvement |
| CPU architecture | Nvidia’s GPU platforms and associated CPU offerings | RISC-V technology integrated into its platform |
| Scaling | Specialized interconnects and established data-center systems | Ethernet-based networking and scalable system designs |
| Memory and systems | Broad range of GPU memory and server configurations | Different local and attached-memory configurations depending on the product |
| Customer trade-off | Higher ecosystem maturity and usually lower migration risk | Potential flexibility or cost advantages, but a smaller ecosystem and greater validation burden |
There is no universal winner. Performance and economics depend on the exact model, numerical precision, batch size, latency target, concurrency, memory requirements, networking, power, software version, and comparison system.
What will the funding pay for?
Tenstorrent said the new capital would support:
- Expansion of its engineering organization.
- Strengthening its global supply chain.
- Development of large AI training servers to demonstrate its technology.
- Continued chip and system development.
- Growth of its hardware, software, and developer ecosystem.
That is a meaningful use of capital for a semiconductor startup. Designing and validating chips requires expensive engineering, verification, packaging, fabrication, boards, servers, compilers, kernels, documentation, customer support, and production testing.
Rank #3
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
However, funding is not revenue. It does not prove that products will achieve a particular performance level, that manufacturing capacity is secured, or that customers will deploy the systems at scale. Tenstorrent also said it had more than $150 million in commercial contracts at the time of the announcement. That is a company-reported contract figure, not the same as recognized revenue, shipments, profit, or market share.
What has Tenstorrent delivered since the financing?
As of August 2026, Tenstorrent has moved beyond a purely speculative startup story. Its official product pages list developer cards and Galaxy systems based on its Blackhole and Wormhole platforms.
Listed prices included:
- Blackhole p100a: starting at $999.
- Blackhole p150a/p150b: listed at $1,399.
- Galaxy Wormhole: from $70,000.
- Galaxy Blackhole: from $110,000.
- Blackhole supercluster: from $440,000.
These are official list or starting prices checked in August 2026, not complete production-cluster costs. Enterprise buyers may also need networking, installation, support, storage, datacenter power, cooling, and model-optimization work. Larger configurations require sales contact or configuration rather than a simple online checkout.
Tenstorrent’s Galaxy Blackhole page lists a configuration with 32 Blackhole ASICs, 1 TB of GDDR6 memory, and 23 PFLOPS of Block FP8 performance. Those are vendor specifications and should not be treated as independently verified performance for every workload.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Tenstorrent announced general availability of Galaxy Blackhole on April 28, 2026, and its newsroom lists developments including product launches, partnerships, and cloud availability through Koyeb. The company’s April 2026 performance announcement also reported results for video generation and large-language-model serving, including a Prodia collaboration. Those results should be evaluated using the published model, precision, software version, comparison hardware, and measurement method.
See the official card listings, Galaxy systems page, and performance and availability announcement for current product details.
Rank #4
- 48GB AI graphics accelerator
Can Tenstorrent really compete with Nvidia?
Yes, as a credible alternative in selected workloads; no, not yet as an Nvidia-equivalent business.
Tenstorrent now has funding, products, a software strategy, and commercial activity. That makes it more than a concept company. But Nvidia’s position is defined by global scale, a deeply established developer ecosystem, broad cloud and server availability, and years of customer adoption.
Tenstorrent may be attractive when:
- The workload is inference-heavy and supported by its software stack.
- The customer wants an alternative to CUDA.
- The organization values access to hardware and compiler internals.
- Ethernet-based scaling fits its existing networking expertise.
- The buyer can benchmark and optimize its own models.
- Lower acquisition or operating cost matters more than maximum ecosystem maturity.
Nvidia is probably the safer choice when:
- The workload depends on CUDA-specific libraries or proprietary kernels.
- The team needs broad framework compatibility immediately.
- The organization lacks engineers to port and optimize models.
- Training reliability and third-party support outweigh hardware price.
- The buyer needs a large, proven supplier with extensive cloud availability.
The risks buyers should evaluate
Software maturity
An open-source-oriented stack can improve transparency and portability, but it can also shift work to the customer. Before buying, verify support for the required frameworks, model architectures, operators, quantization formats, distributed-training features, and serving tools. Tenstorrent provides developer documentation, but documentation breadth is not the same as independently verified production readiness for every model.
Performance claims
“Industry-leading” is not a universal result. A system can perform well for one inference model and poorly for another. Ask for results on the exact workload, including prompt length, generation length, batch size, concurrency, precision, memory configuration, software version, power draw, and comparison hardware.
Training versus inference
A product that is economically attractive for inference may not be the best choice for large-scale training. Training places different demands on memory capacity, communication, checkpointing, compiler support, and cluster reliability.
Supply and manufacturing
Earlier coverage described plans involving TSMC and Samsung and exploration of advanced process nodes. Those statements should be understood as reported plans or roadmaps, not proof that every Tenstorrent product is manufactured by both companies or that a particular 2-nanometer product has reached production.
Best Value
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
Total cost
A $999 developer card is not a one-for-one replacement for a datacenter GPU. Production economics include hardware, host systems, networking, power, cooling, support, software engineering, downtime, and the cost of migrating models.
Questions to ask before buying
- Does the quoted price cover only hardware, or also networking, support, installation, and software?
- What is the delivered performance on the organization’s own model?
- What are the power, cooling, and rack requirements?
- Are all required operators, quantization formats, and frameworks supported?
- How much engineering time is needed to move from CUDA?
- What warranty, replacement, and long-term support arrangements apply?
- Is the product available in the buyer’s country and on the required schedule?
- Can the hardware be rented or trialed before purchase?
- What cloud instances are available, and what are their current prices?
- How does the complete cost compare with Nvidia, AMD, cloud GPUs, or specialized inference providers?
Tenstorrent lists documentation and support resources at tenstorrent.com/support. Its newsroom has also described Wormhole availability through Koyeb Cloud; current geography, specifications, and pricing should be confirmed before committing.
What this means for investors and finance readers
The financing is evidence of substantial investor interest in alternatives to Nvidia’s dominance. The investor list also suggests that the opportunity is strategically relevant to financial firms, industrial companies, automotive interests, and technology investors.
It is not, by itself, evidence that Tenstorrent is profitable, that the round will generate a particular return, or that Nvidia’s market position is about to collapse. Private-company valuations are not the same as public-market prices, and a Series D investment can still carry significant technology, execution, manufacturing, customer-concentration, and future-funding risk.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallFor anyone evaluating the company as an investment opportunity, the important questions are commercialization, recurring revenue, gross margins, production volume, customer retention, cash burn, future capital needs, and the cost of competing with Nvidia and other accelerator vendors. The funding headline answers none of those questions on its own.
Bottom line
Tenstorrent’s December 2024 Series D was a serious financing event: more than $693 million led by Samsung Securities and AFW Partners, with Bezos Expeditions and a broad group of investors participating. The capital has helped the company progress toward commercial hardware, including developer cards and Galaxy systems.
But the deal should be described accurately. Bezos Expeditions—not a disclosed personal cheque from Jeff Bezos—participated; Samsung Securities—not necessarily Samsung Electronics—was a lead investor; and $700 million is a rounded version of the official figure. Tenstorrent is a credible, well-funded Nvidia challenger, but it remains an alternative platform whose value depends on workload-specific performance, software support, delivered cost, and customer willingness to accept more ecosystem risk.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →
