Blueshift Memory Limited

Blueshift Memory is ​a Cambridge UK-based technology company that addresses the von Neumann bottleneck (or tailback) in high-performance computing​ to enable processing to keep pace with spiralling data demands.​ Our new proprietary memory chip design optimizes the memory architecture so that large data sets and time-critical data can be more efficiently handled, to ​deliver exceptional memory access speed​ up to 1,000 times faster for specific data-focused applications. ​

Our team analysed and categorised thousands of algorithms used by data processing applications to discover a general technique that can adapt the memory retrieval to fit the data access patterns, yielding benefits in retrieval performance and energy consumption.
BSM memory is optimised for Big Data and time critical solutions. The base performance boost is 5-20x for common sequential algorithms. For specific use-cases, such as Online Transaction Processing, there are additional benefits that BSM memory enables to yield 100-1000x speed-ups.

The size of the DRAM market exceeds $80+ billion per year, about 95% of which is split between three companies.
The global DRAM market is expected to grow at a CAGR of 28.70% fr om 2018–2023, based on the Market Research Future 2023 forecast report.
Increased performance smartphones, virtual reality, Internet of Things (IoT) devices and autonomous vehicles are all expanding markets for the semiconductor industry. All of these use dynamic data calculation wh ere Blueshift Memory architecture has immense advantages.

In today’s multi-core System on a Chip (SoC) designs, the existing memory technologies, such as HBM and DDR4/5 are not able to satisfy the data demands of the CPU cores. This is commonly referred to as the “Memory Wall”, which has progressively worsened over time. CPUs and non-volatile data storage went through a huge evolution in previous decades, but memory still uses the same basic architecture. There have been attempts to increase the number of memory buses and the frequency, to redesign (HBM) or to optimise the L1-L3 cache usage but the main structure remains the same as it has been since the 1960’s.

Our team analysed and categorised thousands of algorithms used by data processing applications to discover a general technique that can adapt the memory retrieval to fit the data access patterns, yielding benefits in retrieval performance and energy consumption. The technique is independent of the memory cell technology (DRAM, SRAM, MRAM, FeRAM etc.) and scale of the memory, providing benefits for servers, desktops, laptops, and mobile phones.
Blueshift Memory also provides a software runtime to enable a more efficient software development platform. The runtime provides additional value-add by managing the Blueshift Memory modules to further improve memory retrieval performance.
BSM memory is optimised for Big Data and time critical solutions. The base performance boost is 5-20x for common sequential algorithms. For specific use-cases, such as Online Transaction Processing, there are additional benefits that BSM memory enables to yield 100-1000x speed-ups.

Samsung, SK Hynix, Micron

BSM memory is optimised for Big Data and time critical solutions. The base performance boost is 5-20x for common sequential algorithms. For specific use-cases, such as Online Transaction Processing, there are additional benefits that BSM memory enables to yield 100-1000x speed-ups.

BSM solutions have been proven to deliver significantly improved memory access performance in the following fields of computing:
● Nuclear and pharmaceutical research,
● Live Retail management,
● FinTech,
● Missile Technology,
● Cloud and High-Performance Computing,
● Image Processing
● Smart Cities and
● Meteorology-Climate research.

Fundraising strategy
1. Seed investment to fully fund start-up development and MVP
• £1.5 Million
2. Create the product ASIC prototype
• $3 Million
3. Scaling through SoC services and further develop technology licensing portfolio
• $5 Million

Based on the Cambridge University Judge Business School study, even with the most conservative estimates, the forecasted revenue for Blueshift Memory should be expected to reach >$30-100M in 5 years.

Main phases:
1. BSM is an IP block business with IP blocks and IP block generators
2. MVP R&D - SoC based product development (revenue generating phase)
3. Scaling through SoC services
4. IP blocks get integrated into the FPGA and ASIC design, Licensing of Chip and start of
mass production manufacturing through strategic semiconductor partners
The short-term target​ of the Blueshift Memory is to create a Minimum Viable Product (MVP) and start to provide our System On a Chip (SoC) based services for specific use-cases:
● Big data, High Frequency Trading (HFT),
● Autonomous vehicles,
● AI acceleration,
● Blockchain,
● Cryptography,
● eDiscovery,
● AR/VR,
● Image processing and Compression.
BSM is currently negotiating SoC collaborations on these fields with market leading companies.
Our long-term sales strategy​ is to licence our technology through NRE collaborations and start the mass production of our memory chip as a fabless company.

MVP, Startup development

20% equity for £1.5Million

Blueshift Memory is a fabless IP semiconductor business designed to integrate into the FPGA and ASIC design flow, with a software runtime to make system integration into full SoC-based systems easier, faster and more reliable​.

YCombinator

• Finalist (Best 7) in Big Data/AI category in Hello Tomorrow 2019 Worldwide
Challenge (4,500 startups from 120 countries)
• Finalist in Cambridge Independent Science and Technology Awards 2019 in two
categories:​‘Startup of the Year’​, and ‘​The One to Watch​’

• FPGA based prototype
• 2 patents
• Tested over 2000+ application/project scenarios