the United States’s
ai infrastructure industry,
in focus.

Memory Bottlenecks in AI Hardware

39.3% of this theme group

Researchers highlight limits of von Neumann designs and robustness issues in in‑memory processing for machine learning.

Bottleneck (3) · In-Memory Processing (3) · Von Neumann architecture (2) · Robustness (2)

In-Memory Computing in Emerging Memory Technologies for Machine Learning: An Overview

Resistive-RAM-Based In-Memory Computing for Neural Network: A Review

Photonics and Edge Architecture Challenges

39.3% of this theme group

Papers point to integration difficulties of photonic devices and edge tensor units within existing computer architectures.

Photonics (2) · Embedded system (2) · Computer architecture (2)

Emerging devices and packaging strategies for electronic-photonic AI accelerators: opinion

Accelerating applications using edge tensor processing units

Magnetoresistive Memory Constraints

14.3% of this theme group

Studies discuss design and optimization hurdles for MRAM‑based in‑memory computing solutions.

Magnetoresistive random-access memory (2)

Domain-Specific STT-MRAM-Based In-Memory Computing: A Survey

Modeling and Optimization of SRAM-based In-Memory Computing Hardware Design

Variability in Resistive Memory Devices

42.9% of this theme group

Authors caution about conductance drift and interconnect reliability when using RRAM and memristor arrays for inference.

Resistive random-access memory (7) · Spice (2) · Memristor (2) · Interconnection (2)

OxRRAM-Based Analog in-Memory Computing for Deep Neural Network Inference: A Conductance Variability Study

Characterization and Mitigation of Relaxation Effects on Multi-level RRAM based In-Memory Computing

Energy Efficiency and Precision Limits

33.3% of this theme group

Works warn of power‑intensive operation and bit‑slicing accuracy concerns in SRAM and phase‑change memory in‑memory schemes.

Random access memory (2) · Efficient energy use (2) · Static random-access memory (2)

In-Memory Computing: Advances and Prospects

Precision of bit slicing with in-memory computing based on analog phase-change memory crossbars

Tensor Unit Resilience in Harsh Environments

14.3% of this theme group

Research flags soft‑error susceptibility and timing‑error risks for tensor processors used in low‑power or space applications.

Tensor (2) · Resilience (2) · Power (2)

Soft-Error Characterization and Mitigation Strategies for Edge Tensor Processing Units in Space

STRIVE: Enabling Choke Point Detection and Timing Error Resilience in a Low-Power Tensor Processing Unit

Scalable Parallel In‑Memory Accelerators

23.6% of this theme group

Large body of work explores programmable, high‑throughput architectures that combine analog and digital in‑memory computing.

Parallel computing (41) · Computer architecture (22) · Scalability (16) · Electronic engineering (9)

Scalable and Programmable Neural Network Inference Accelerator Based on In-Memory Computing

Hybrid Analog-Digital In-Memory Computing

FPGA‑Based Deep Learning Platforms

19.8% of this theme group

Papers present design‑space exploration and open‑source frameworks for efficient, scalable AI accelerators on reconfigurable hardware.

Field-programmable gate array (14) · Computer architecture (20) · Design space exploration (6) · Scalability (12)

FIdelity: Efficient Resilience Analysis Framework for Deep Learning Accelerators

T-DLA: An Open-source Deep Learning Accelerator for Ternarized DNN Models on Embedded FPGA

Hybrid CMOS‑RRAM In‑Memory Systems

18.2% of this theme group

Studies demonstrate monolithic integration of resistive and conventional memories to boost deep‑learning performance.

Static random-access memory (28) · Resistive random-access memory (33) · Magnetoresistive random-access memory (20) · CMOS (10)

High-Throughput In-Memory Computing for Binary Deep Neural Networks With Monolithically Integrated RRAM and 90-nm CMOS

Monolithically Integrated RRAM- and CMOS-Based In-Memory Computing Optimizations for Efficient Deep Learning

Ferroelectric Transistor Computing

16.7% of this theme group

Research highlights ferroelectric FETs and memristors as promising in‑memory elements for few‑shot and other learning tasks.

Ferroelectricity (7) · Transistor (8) · Von Neumann architecture (8) · Memristor (7)

Ferroelectric FET Based In-Memory Computing for Few-Shot Learning

Ferroelectric capacitors and field-effect transistors as in-memory computing elements for machine learning workloads

Reconfigurable Crossbar Computing

8.5% of this theme group

Works investigate resistive crossbar arrays for Boolean function implementation and distributed in‑memory processing.

Crossbar switch (33) · Resistive random-access memory (13) · Boolean function (5) · In-Memory Processing (9)

Distributed In-Memory Computing on Binary RRAM Crossbar

Reconfigurable in-memory computing with resistive memory crossbar

Tensor Processing in Cloud Environments

7.5% of this theme group

Papers analyze performance modeling and datacenter deployment of tensor units for large‑scale computational science.

Tensor (12) · Computational science (7) · Discretization (4) · Cloud computing (5)

In-Datacenter Performance Analysis of a Tensor Processing Unit

A Learned Performance Model for Tensor Processing Units

Fault‑Criticality Analysis for AI Chips

5.7% of this theme group

Research introduces neural‑twin and testing methods to locate and mitigate faults in AI accelerator designs.

Criticality (7) · Fault (4) · Automatic test pattern generation (4) · Reliability engineering (3)

Efficient Fault-Criticality Analysis for AI Accelerators using a Neural Twin

C-Testing and Efficient Fault Localization for AI Accelerators

San Francisco

39.1% across the displayed hubs

3 companies hiring · $869.7M disclosed funding

New York City

14.8% across the displayed hubs

2 companies hiring · $1.1B disclosed funding

San Jose

10.4% across the displayed hubs

4 companies hiring · $2.3B disclosed funding

Austin

8.7% across the displayed hubs

2 companies hiring · $419.6M disclosed funding

Santa Clara

7.0% across the displayed hubs

3 companies hiring · $1.9B disclosed funding

Mountain View

7.0% across the displayed hubs

1 companies hiring · $161.9M disclosed funding

Palo Alto

7.0% across the displayed hubs

0 companies hiring · $1.5B disclosed funding

Miami

6.1% across the displayed hubs

1 companies hiring · $16.7M disclosed funding