diff --git a/src/stm32/Kconfig b/src/stm32/Kconfig
index b9aac0446..aa67dc91e 100644
--- a/src/stm32/Kconfig
+++ b/src/stm32/Kconfig
@@ -13,7 +13,7 @@ config STM32_SELECT
     select HAVE_GPIO_HARD_PWM if MACH_STM32F070 || MACH_STM32F072 || MACH_STM32F1 || MACH_STM32F4 || MACH_STM32F7 || MACH_STM32G0 || MACH_STM32H7
     select HAVE_STRICT_TIMING
     select HAVE_CHIPID
-    select HAVE_STEPPER_OPTIMIZED_BOTH_EDGE
+    select HAVE_STEPPER_OPTIMIZED_BOTH_EDGE if !MACH_STM32H7
     select HAVE_BOOTLOADER_REQUEST
     select HAVE_LIMITED_CODE_SIZE if FLASH_SIZE < 0x10000
 
diff --git a/src/stm32/Makefile b/src/stm32/Makefile
index 1ce30c519..0ee71dcf1 100644
--- a/src/stm32/Makefile
+++ b/src/stm32/Makefile
@@ -37,7 +37,7 @@ CFLAGS_klipper.elf += -T $(OUT)src/generic/armcm_link.ld
 $(OUT)klipper.elf: $(OUT)src/generic/armcm_link.ld
 
 # Add source files
-src-y += stm32/watchdog.c stm32/gpio.c stm32/clockline.c stm32/dfu_reboot.c
+src-y += stm32/watchdog.c stm32/clockline.c stm32/dfu_reboot.c
 src-y += generic/crc16_ccitt.c
 src-y += generic/armcm_boot.c generic/armcm_irq.c generic/armcm_reset.c
 src-$(CONFIG_MACH_STM32F0) += stm32/stm32f0.c ../lib/stm32f0/system_stm32f0xx.c
@@ -52,8 +52,9 @@ src-$(CONFIG_MACH_STM32L4) += stm32/stm32l4.c ../lib/stm32l4/system_stm32l4xx.c
 timer-src-y := generic/armcm_timer.c
 timer-src-$(CONFIG_MACH_STM32F0) := generic/timer_irq.c stm32/stm32f0_timer.c
 timer-src-$(CONFIG_MACH_STM32G0) := generic/timer_irq.c stm32/stm32f0_timer.c
-gpio-src-y := stm32/gpioperiph.c
-gpio-src-$(CONFIG_MACH_STM32F1) :=
+gpio-src-y := stm32/gpio.c stm32/gpioperiph.c
+gpio-src-$(CONFIG_MACH_STM32F1) := stm32/gpio.c
+gpio-src-$(CONFIG_MACH_STM32H7) := stm32/stm32h7_gpio.c stm32/gpioperiph.c
 src-y += $(timer-src-y) $(gpio-src-y)
 adc-src-y := stm32/adc.c
 adc-src-$(CONFIG_MACH_STM32F0) := stm32/stm32f0_adc.c
diff --git a/src/stm32/gpio.h b/src/stm32/gpio.h
index 6a5f910cc..0bfffc4b4 100644
--- a/src/stm32/gpio.h
+++ b/src/stm32/gpio.h
@@ -5,7 +5,10 @@
 
 struct gpio_out {
     void *regs;
-    uint32_t bit;
+    union {
+        struct odr_cache *oc;  // stm32h7 uses 'oc'; all others use 'bit'
+        uint32_t bit;
+    };
 };
 struct gpio_out gpio_out_setup(uint32_t pin, uint32_t val);
 void gpio_out_reset(struct gpio_out g, uint32_t val);
diff --git a/src/stm32/stm32h7_gpio.c b/src/stm32/stm32h7_gpio.c
new file mode 100644
index 000000000..f653c5708
--- /dev/null
+++ b/src/stm32/stm32h7_gpio.c
@@ -0,0 +1,167 @@
+// GPIO functions optimized for stm32h7
+//
+// Copyright (C) 2019-2025  Kevin O'Connor <kevin@koconnor.net>
+//
+// This file may be distributed under the terms of the GNU GPLv3 license.
+
+#include <string.h> // ffs
+#include "board/irq.h" // irq_save
+#include "command.h" // DECL_ENUMERATION_RANGE
+#include "gpio.h" // gpio_out_setup
+#include "internal.h" // gpio_peripheral
+#include "sched.h" // sched_shutdown
+
+DECL_ENUMERATION_RANGE("pin", "PA0", GPIO('A', 0), 16);
+DECL_ENUMERATION_RANGE("pin", "PB0", GPIO('B', 0), 16);
+DECL_ENUMERATION_RANGE("pin", "PC0", GPIO('C', 0), 16);
+#ifdef GPIOD
+DECL_ENUMERATION_RANGE("pin", "PD0", GPIO('D', 0), 16);
+#endif
+#ifdef GPIOE
+DECL_ENUMERATION_RANGE("pin", "PE0", GPIO('E', 0), 16);
+#endif
+#ifdef GPIOF
+DECL_ENUMERATION_RANGE("pin", "PF0", GPIO('F', 0), 16);
+#endif
+#ifdef GPIOG
+DECL_ENUMERATION_RANGE("pin", "PG0", GPIO('G', 0), 16);
+#endif
+#ifdef GPIOH
+DECL_ENUMERATION_RANGE("pin", "PH0", GPIO('H', 0), 16);
+#endif
+#ifdef GPIOI
+DECL_ENUMERATION_RANGE("pin", "PI0", GPIO('I', 0), 16);
+#endif
+
+GPIO_TypeDef * const digital_regs[] = {
+    ['A' - 'A'] = GPIOA, GPIOB, GPIOC,
+#ifdef GPIOD
+    ['D' - 'A'] = GPIOD,
+#endif
+#ifdef GPIOE
+    ['E' - 'A'] = GPIOE,
+#endif
+#ifdef GPIOF
+    ['F' - 'A'] = GPIOF,
+#endif
+#ifdef GPIOG
+    ['G' - 'A'] = GPIOG,
+#endif
+#ifdef GPIOH
+    ['H' - 'A'] = GPIOH,
+#endif
+#ifdef GPIOI
+    ['I' - 'A'] = GPIOI,
+#endif
+};
+
+// Convert a register and bit location back to an integer pin identifier
+static int
+regs_to_pin(GPIO_TypeDef *regs, uint32_t bit)
+{
+    int i;
+    for (i=0; i<ARRAY_SIZE(digital_regs); i++)
+        if (digital_regs[i] == regs)
+            return GPIO('A' + i, ffs(bit)-1);
+    return 0;
+}
+
+// Verify that a gpio is a valid pin
+static int
+gpio_valid(uint32_t pin)
+{
+    uint32_t port = GPIO2PORT(pin);
+    return port < ARRAY_SIZE(digital_regs) && digital_regs[port];
+}
+
+// The stm32h7 has very slow read access to the gpio registers.  Cache
+// the ODR register in memory to speed up toggle operations.
+static struct odr_cache {
+    uint32_t bsrr;
+} ODR_CACHE[ARRAY_SIZE(digital_regs) * 16] __aligned(64);
+
+struct gpio_out
+gpio_out_setup(uint32_t pin, uint32_t val)
+{
+    if (!gpio_valid(pin))
+        shutdown("Not an output pin");
+    uint32_t port = GPIO2PORT(pin);
+    GPIO_TypeDef *regs = digital_regs[port];
+    gpio_clock_enable(regs);
+
+    struct gpio_out g = { .regs=regs, .oc=&ODR_CACHE[pin] };
+    gpio_out_reset(g, val);
+    return g;
+}
+
+void
+gpio_out_reset(struct gpio_out g, uint32_t val)
+{
+    GPIO_TypeDef *regs = g.regs;
+    uint32_t pin = g.oc - ODR_CACHE;
+    uint32_t bit_num = ((uint32_t)g.oc / sizeof(*g.oc)) % 16;
+    uint32_t bsrr = 1 << (val ? bit_num : bit_num + 16);
+    irqstatus_t flag = irq_save();
+    g.oc->bsrr = bsrr;
+    regs->BSRR = bsrr;
+    gpio_peripheral(pin, GPIO_OUTPUT, 0);
+    irq_restore(flag);
+}
+
+void
+gpio_out_toggle_noirq(struct gpio_out g)
+{
+    GPIO_TypeDef *regs = g.regs;
+    uint32_t bsrr = (g.oc->bsrr << 16) | (g.oc->bsrr >> 16);
+    g.oc->bsrr = bsrr;
+    regs->BSRR = bsrr;
+}
+
+void
+gpio_out_toggle(struct gpio_out g)
+{
+    irqstatus_t flag = irq_save();
+    gpio_out_toggle_noirq(g);
+    irq_restore(flag);
+}
+
+void
+gpio_out_write(struct gpio_out g, uint32_t val)
+{
+    GPIO_TypeDef *regs = g.regs;
+    uint32_t bit_num = ((uint32_t)g.oc / sizeof(*g.oc)) % 16;
+    uint32_t bsrr = 1 << (val ? bit_num : bit_num + 16);
+    irqstatus_t flag = irq_save();
+    g.oc->bsrr = bsrr;
+    regs->BSRR = bsrr;
+    irq_restore(flag);
+}
+
+
+struct gpio_in
+gpio_in_setup(uint32_t pin, int32_t pull_up)
+{
+    if (!gpio_valid(pin))
+        shutdown("Not a valid input pin");
+    GPIO_TypeDef *regs = digital_regs[GPIO2PORT(pin)];
+    struct gpio_in g = { .regs=regs, .bit=GPIO2BIT(pin) };
+    gpio_in_reset(g, pull_up);
+    return g;
+}
+
+void
+gpio_in_reset(struct gpio_in g, int32_t pull_up)
+{
+    GPIO_TypeDef *regs = g.regs;
+    int pin = regs_to_pin(regs, g.bit);
+    irqstatus_t flag = irq_save();
+    gpio_peripheral(pin, GPIO_INPUT, pull_up);
+    irq_restore(flag);
+}
+
+uint8_t
+gpio_in_read(struct gpio_in g)
+{
+    GPIO_TypeDef *regs = g.regs;
+    return !!(regs->IDR & g.bit);
+}